draft-ietf-tsvwg-byte-pkt-congest-04.txt   draft-ietf-tsvwg-byte-pkt-congest-05.txt 
Transport Area Working Group B. Briscoe Transport Area Working Group B. Briscoe
Internet-Draft BT Internet-Draft BT
Updates: 2309 (if approved) J. Manner Updates: 2309 (if approved) J. Manner
Intended status: Informational Aalto University Intended status: BCP Aalto University
Expires: September 15, 2011 March 14, 2011 Expires: May 3, 2012 October 31, 2011
Byte and Packet Congestion Notification Byte and Packet Congestion Notification
draft-ietf-tsvwg-byte-pkt-congest-04 draft-ietf-tsvwg-byte-pkt-congest-05
Abstract Abstract
This memo concerns dropping or marking packets using active queue This memo concerns dropping or marking packets using active queue
management (AQM) such as random early detection (RED) or pre- management (AQM) such as random early detection (RED) or pre-
congestion notification (PCN). We give three strong recommendations: congestion notification (PCN). We give three strong recommendations:
(1) packet size should be taken into account when transports read (1) packet size should be taken into account when transports read and
congestion indications, (2) packet size should not be taken into respond to congestion indications, (2) packet size should not be
account when network equipment creates congestion signals (marking, taken into account when network equipment creates congestion signals
dropping), and therefore (3) the byte-mode packet drop variant of the (marking, dropping), and therefore (3) the byte-mode packet drop
RED AQM algorithm that drops fewer small packets should not be used. variant of the RED AQM algorithm that drops fewer small packets
should not be used.
Status of This Memo Status of This Memo
This Internet-Draft is submitted in full conformance with the This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79. provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet- working documents as Internet-Drafts. The list of current Internet-
Drafts is at http://datatracker.ietf.org/drafts/current/. Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress." material or to cite them other than as "work in progress."
This Internet-Draft will expire on September 15, 2011. This Internet-Draft will expire on May 3, 2012.
Copyright Notice Copyright Notice
Copyright (c) 2011 IETF Trust and the persons identified as the Copyright (c) 2011 IETF Trust and the persons identified as the
document authors. All rights reserved. document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents Provisions Relating to IETF Documents
(http://trustee.ietf.org/license-info) in effect on the date of (http://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License. described in the Simplified BSD License.
Table of Contents Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 5 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.1. Terminology and Scoping . . . . . . . . . . . . . . . . . 7 1.1. Terminology and Scoping . . . . . . . . . . . . . . . . . 6
1.2. Example Comparing Packet-Mode Drop and Byte-Mode Drop . . 7
2. Recommendations . . . . . . . . . . . . . . . . . . . . . . . 8 2. Recommendations . . . . . . . . . . . . . . . . . . . . . . . 8
2.1. Recommendation on Queue Measurement . . . . . . . . . . . 8 2.1. Recommendation on Queue Measurement . . . . . . . . . . . 9
2.2. Recommendation on Notifying Congestion . . . . . . . . . . 9 2.2. Recommendation on Encoding Congestion Notification . . . . 9
2.3. Recommendation on Responding to Congestion . . . . . . . . 10 2.3. Recommendation on Responding to Congestion . . . . . . . . 10
2.4. Recommendation on Handling Congestion Indications when 2.4. Recommendation on Handling Congestion Indications when
Splitting or Merging Packets . . . . . . . . . . . . . . . 11 Splitting or Merging Packets . . . . . . . . . . . . . . . 11
3. Motivating Arguments . . . . . . . . . . . . . . . . . . . . . 11 3. Motivating Arguments . . . . . . . . . . . . . . . . . . . . . 11
3.1. Scaling Congestion Control with Packet Size . . . . . . . 11 3.1. Avoiding Perverse Incentives to (Ab)use Smaller Packets . 12
3.2. Transport-Independent Network . . . . . . . . . . . . . . 12 3.2. Small != Control . . . . . . . . . . . . . . . . . . . . . 13
3.3. Avoiding Perverse Incentives to (Ab)use Smaller Packets . 13 3.3. Transport-Independent Network . . . . . . . . . . . . . . 13
3.4. Small != Control . . . . . . . . . . . . . . . . . . . . . 14 3.4. Scaling Congestion Control with Packet Size . . . . . . . 14
3.5. Implementation Efficiency . . . . . . . . . . . . . . . . 14 3.5. Implementation Efficiency . . . . . . . . . . . . . . . . 16
3.6. Why now? . . . . . . . . . . . . . . . . . . . . . . . . . 14 4. A Survey and Critique of Past Advice . . . . . . . . . . . . . 16
4. A Survey and Critique of Past Advice . . . . . . . . . . . . . 15
4.1. Congestion Measurement Advice . . . . . . . . . . . . . . 16 4.1. Congestion Measurement Advice . . . . . . . . . . . . . . 16
4.1.1. Fixed Size Packet Buffers . . . . . . . . . . . . . . 17 4.1.1. Fixed Size Packet Buffers . . . . . . . . . . . . . . 17
4.1.2. Congestion Measurement without a Queue . . . . . . . . 18 4.1.2. Congestion Measurement without a Queue . . . . . . . . 18
4.2. Congestion Notification Advice . . . . . . . . . . . . . . 18 4.2. Congestion Notification Advice . . . . . . . . . . . . . . 19
4.2.1. Network Bias when Encoding . . . . . . . . . . . . . . 18 4.2.1. Network Bias when Encoding . . . . . . . . . . . . . . 19
4.2.2. Transport Bias when Decoding . . . . . . . . . . . . . 20 4.2.2. Transport Bias when Decoding . . . . . . . . . . . . . 21
4.2.3. Making Transports Robust against Control Packet 4.2.3. Making Transports Robust against Control Packet
Losses . . . . . . . . . . . . . . . . . . . . . . . . 22 Losses . . . . . . . . . . . . . . . . . . . . . . . . 22
4.2.4. Congestion Notification: Summary of Conflicting 4.2.4. Congestion Notification: Summary of Conflicting
Advice . . . . . . . . . . . . . . . . . . . . . . . . 22 Advice . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2.5. RED Implementation Status . . . . . . . . . . . . . . 23
5. Outstanding Issues and Next Steps . . . . . . . . . . . . . . 24 5. Outstanding Issues and Next Steps . . . . . . . . . . . . . . 24
5.1. Bit-congestible World . . . . . . . . . . . . . . . . . . 24 5.1. Bit-congestible Network . . . . . . . . . . . . . . . . . 24
5.2. Bit- & Packet-congestible World . . . . . . . . . . . . . 25 5.2. Bit- & Packet-congestible Network . . . . . . . . . . . . 24
6. Security Considerations . . . . . . . . . . . . . . . . . . . 26 6. Security Considerations . . . . . . . . . . . . . . . . . . . 24
7. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 27 7. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 25
8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 28 8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 26
9. Comments Solicited . . . . . . . . . . . . . . . . . . . . . . 28 9. Comments Solicited . . . . . . . . . . . . . . . . . . . . . . 27
10. References . . . . . . . . . . . . . . . . . . . . . . . . . . 28 10. References . . . . . . . . . . . . . . . . . . . . . . . . . . 27
10.1. Normative References . . . . . . . . . . . . . . . . . . . 28 10.1. Normative References . . . . . . . . . . . . . . . . . . . 27
10.2. Informative References . . . . . . . . . . . . . . . . . . 29 10.2. Informative References . . . . . . . . . . . . . . . . . . 27
Appendix A. Idealised Wire Protocol . . . . . . . . . . . . . . . 33 Appendix A. Survey of RED Implementation Status . . . . . . . . . 31
A.1. Protocol Coding . . . . . . . . . . . . . . . . . . . . . 33 Appendix B. Sufficiency of Packet-Mode Drop . . . . . . . . . . . 32
A.2. Example Scenarios . . . . . . . . . . . . . . . . . . . . 34 B.1. Packet-Size (In)Dependence in Transports . . . . . . . . . 33
A.2.1. Notation . . . . . . . . . . . . . . . . . . . . . . . 34 B.2. Bit-Congestible and Packet-Congestible Indications . . . . 36
A.2.2. Bit-congestible resource, equal bit rates (Ai) . . . . 35
A.2.3. Bit-congestible resource, equal packet rates (Bi) . . 36 Appendix C. Byte-mode Drop Complicates Policing Congestion
A.2.4. Pkt-congestible resource, equal bit rates (Aii) . . . 37 Response . . . . . . . . . . . . . . . . . . . . . . 37
A.2.5. Pkt-congestible resource, equal packet rates (Bii) . . 37 Appendix D. Changes from Previous Versions . . . . . . . . . . . 38
Appendix B. Byte-mode Drop Complicates Policing Congestion
Response . . . . . . . . . . . . . . . . . . . . . . 38
Appendix C. Changes from Previous Versions . . . . . . . . . . . 39
1. Introduction 1. Introduction
This memo is initially concerned with how we should correctly scale This memo concerns how we should correctly scale congestion control
congestion control functions with packet size for the long term. But functions with packet size for the long term. It also recognises
it also recognises that expediency may be necessary to deal with that expediency may be necessary to deal with existing widely
existing widely deployed protocols that don't live up to the long deployed protocols that don't live up to the long term goal.
term goal.
When notifying congestion, the problem of how (and whether) to take When notifying congestion, the problem of how (and whether) to take
packet sizes into account has exercised the minds of researchers and packet sizes into account has exercised the minds of researchers and
practitioners for as long as active queue management (AQM) has been practitioners for as long as active queue management (AQM) has been
discussed. Indeed, one reason AQM was originally introduced was to discussed. Indeed, one reason AQM was originally introduced was to
reduce the lock-out effects that small packets can have on large reduce the lock-out effects that small packets can have on large
packets in drop-tail queues. This memo aims to state the principles packets in drop-tail queues. This memo aims to state the principles
we should be using and to come to conclusions on what these we should be using and to outline how these principles will affect
principles will mean for future protocol design, taking into account future protocol design, taking into account the existing deployments
the deployments we have already. we have already.
The byte vs. packet dilemma arises at three stages in the congestion The question of whether to take into account packet size arises at
notification process: three stages in the congestion notification process:
Measuring congestion: When the congested resource decides locally to Measuring congestion: When a congested resource measures locally how
measure how congested it is, should the queue measure its length congested it is, should it measure its queue length in bytes or
in bytes or packets? packets?
Encoding congestion notification into the wire protocol: When the Encoding congestion notification into the wire protocol: When a
congested network resource decides whether to notify the level of congested network resource notifies its level of congestion,
congestion by dropping or marking a particular packet, should its should it drop / mark each packet dependent on the byte-size of
decision depend on the byte-size of the particular packet being the particular packet in question?
dropped or marked?
Decoding congestion notification from the wire protocol: When the Decoding congestion notification from the wire protocol: When a
transport interprets the notification in order to decide how much transport interprets the notification in order to decide how much
to respond to congestion, should it take into account the byte- to respond to congestion, should it take into account the byte-
size of each missing or marked packet? size of each missing or marked packet?
Consensus has emerged over the years concerning the first stage: Consensus has emerged over the years concerning the first stage:
whether queues are measured in bytes or packets, termed byte-mode whether queues are measured in bytes or packets, termed byte-mode
queue measurement or packet-mode queue measurement. This memo queue measurement or packet-mode queue measurement. Section 2.1 of
records this consensus in the RFC Series. In summary the choice this memo records this consensus in the RFC Series. In summary the
solely depends on whether the resource is congested by bytes or choice solely depends on whether the resource is congested by bytes
packets. or packets.
The controversy is mainly around the last two stages: whether to The controversy is mainly around the last two stages: whether to
allow for the size of the specific packet notifying congestion i) allow for the size of the specific packet notifying congestion i)
when the network encodes or ii) when the transport decodes the when the network encodes or ii) when the transport decodes the
congestion notification. congestion notification.
Currently, the RFC series is silent on this matter other than a paper Currently, the RFC series is silent on this matter other than a paper
trail of advice referenced from [RFC2309], which conditionally trail of advice referenced from [RFC2309], which conditionally
recommends byte-mode (packet-size dependent) drop [pktByteEmail]. recommends byte-mode (packet-size dependent) drop [pktByteEmail].
Reducing drop of small packets certainly has some tempting Reducing drop of small packets certainly has some tempting
advantages: i) it drops less control packets, which tend to be small advantages: i) it drops less control packets, which tend to be small
and ii) it makes TCP's bit-rate less dependent on packet size. and ii) it makes TCP's bit-rate less dependent on packet size.
However, there are ways of addressing these issues at the transport However, there are ways of addressing these issues at the transport
layer, rather than reverse engineering network forwarding to fix the layer, rather than reverse engineering network forwarding to fix the
problems of one specific transport, as byte-mode variant of RED was problems.
designed to do.
The primary purpose of this memo is to build a definitive consensus This memo updates [RFC2309] to deprecate deliberate preferential
against deliberate preferential treatment for small packets in AQM treatment of small packets in AQM algorithms. It recommends that (1)
algorithms and to record this advice within the RFC series. It packet size should be taken into account when transports read
recommends that (1) packet size should be taken into account when congestion indications, (2) not when network equipment writes them.
transports read congestion indications, (2) not when network
equipment writes them.
In particular this means that the byte-mode packet drop variant of In particular this means that the byte-mode packet drop variant of
RED should not be used to drop fewer small packets, because that Random early Detection (RED) should not be used to drop fewer small
creates a perverse incentive for transports to use tiny segments, packets, because that creates a perverse incentive for transports to
consequently also opening up a DoS vulnerability. Fortunately all use tiny segments, consequently also opening up a DoS vulnerability.
the RED implementers who responded to our survey (Section 4.2.4) have Fortunately all the RED implementers who responded to our admittedly
not followed the earlier advice to use byte-mode drop, so the limited survey (Section 4.2.4) have not followed the earlier advice
consensus this memo argues for seems to already exist in to use byte-mode drop, so the position this memo argues for seems to
implementations. already exist in implementations.
However, at the transport layer, TCP congestion control is a widely However, at the transport layer, TCP congestion control is a widely
deployed protocol that doesn't scale correctly with packet size. To deployed protocol that doesn't scale with packet size. To date this
date this hasn't been a significant problem because most TCPs have hasn't been a significant problem because most TCP implementations
been used with similar packet sizes. But, as we design new have been used with similar packet sizes. But, as we design new
congestion controls, we should build in scaling with packet size congestion control mechanisms, the current recommendation is that we
rather than assuming we should follow TCP's example. should build in scaling with packet size rather than assuming we
should follow TCP's example.
This memo continues as follows. First it discusses terminology and This memo continues as follows. First it discusses terminology and
scoping. Section 2 gives the concrete formal recommendations, scoping. Section 2 gives the concrete formal recommendations,
followed by motivating arguments in Section 3. We then critically followed by motivating arguments in Section 3. We then critically
survey the advice given previously in the RFC series and the research survey the advice given previously in the RFC series and the research
literature (Section 4), followed by an assessment of whether or not literature (Section 4), referring to an assessment of whether or not
this advice has been followed in production networks (Section 4.2.5). this advice has been followed in production networks (Appendix A).
To wrap up, outstanding issues are discussed that will need To wrap up, outstanding issues are discussed that will need
resolution both to inform future protocols designs and to handle resolution both to inform future protocol designs and to handle
legacy (Section 5). Then security issues are collected together in legacy (Section 5). Then security issues are collected together in
Section 6 before conclusions are drawn in Section 7. The interested Section 6 before conclusions are drawn in Section 7. The interested
reader can find discussion of more detailed issues on the theme of reader can find discussion of more detailed issues on the theme of
byte vs. packet in the appendices. byte vs. packet in the appendices.
This memo intentionally includes a non-negligible amount of material This memo intentionally includes a non-negligible amount of material
on the subject. A busy reader can jump right into Section 2 to read on the subject. For the busy reader Section 2 summarises the
a summary of the recommendations for the Internet community. recommendations for the Internet community.
1.1. Terminology and Scoping 1.1. Terminology and Scoping
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in [RFC2119]. document are to be interpreted as described in [RFC2119].
Congestion Notification: Rather than aim to achieve what many have Congestion Notification: Congestion notification is a changing
tried and failed, this memo will not try to define congestion. It signal that aims to communicate the probability that the network
will give a working definition of what congestion notification resource(s) will not be able to forward the level of traffic load
should be taken to mean for this document. Congestion offered (or that there is an impending risk that they will not be
notification is a changing signal that aims to communicate the able to).
ratio E/L. E is the instantaneous excess load offered to a
resource that it is either incapable of serving or unwilling to
serve. L is the instantaneous offered load.
The phrase `unwilling to serve' is added, because AQM systems The `impending risk' qualifier is added, because AQM systems (e.g.
(e.g. RED, PCN [RFC5670]) set a virtual limit smaller than the RED, PCN [RFC5670]) set a virtual limit smaller than the actual
actual limit to the resource, then notify when this virtual limit limit to the resource, then notify when this virtual limit is
is exceeded in order to avoid congestion of the actual capacity. exceeded in order to avoid uncontrolled congestion of the actual
capacity.
Note that the denominator is offered load, not capacity. Congestion notification communicates a real number bounded by the
Therefore congestion notification is a real number bounded by the
range [0,1]. This ties in with the most well-understood measure range [0,1]. This ties in with the most well-understood measure
of congestion notification: drop probability (often loosely called of congestion notification: drop probability.
loss rate). It also means that congestion has a natural
interpretation as a probability; the probability of offered
traffic not being served (or being marked as at risk of not being
served).
Explicit and Implicit Notification: The byte vs. packet dilemma Explicit and Implicit Notification: The byte vs. packet dilemma
concerns congestion notification irrespective of whether it is concerns congestion notification irrespective of whether it is
signalled implicitly by drop or using explicit congestion signalled implicitly by drop or using explicit congestion
notification (ECN [RFC3168] or PCN [RFC5670]). Throughout this notification (ECN [RFC3168] or PCN [RFC5670]). Throughout this
document, unless clear from the context, the term marking will be document, unless clear from the context, the term marking will be
used to mean notifying congestion explicitly, while congestion used to mean notifying congestion explicitly, while congestion
notification will be used to mean notifying congestion either notification will be used to mean notifying congestion either
implicitly by drop or explicitly by marking. implicitly by drop or explicitly by marking.
skipping to change at page 8, line 13 skipping to change at page 6, line 52
and firewalls, because load depends on how many packet headers and firewalls, because load depends on how many packet headers
they have to process. Examples of bit-congestible resources are they have to process. Examples of bit-congestible resources are
transmission links, radio power and most buffer memory, because transmission links, radio power and most buffer memory, because
the load depends on how many bits they have to transmit or store. the load depends on how many bits they have to transmit or store.
Some machine architectures use fixed size packet buffers, so Some machine architectures use fixed size packet buffers, so
buffer memory in these cases is packet-congestible (see buffer memory in these cases is packet-congestible (see
Section 4.1.1). Section 4.1.1).
Currently a design goal of network processing equipment such as Currently a design goal of network processing equipment such as
routers and firewalls is to keep packet processing uncongested routers and firewalls is to keep packet processing uncongested
even under worst case bit rates with minimum packet sizes. even under worst case packet rates with runs of minimum size
Therefore, packet-congestion is currently rare [RFC6077; S.3.3], packets. Therefore, packet-congestion is currently rare [RFC6077;
but there is no guarantee that it will not become common with S.3.3], but there is no guarantee that it will not become more
future technology trends. common in future.
Note that information is generally processed or transmitted with a Note that information is generally processed or transmitted with a
minimum granularity greater than a bit (e.g. octets). The minimum granularity greater than a bit (e.g. octets). The
appropriate granularity for the resource in question should be appropriate granularity for the resource in question should be
used, but for the sake of brevity we will talk in terms of bytes used, but for the sake of brevity we will talk in terms of bytes
in this memo. in this memo.
Coarser Granularity: Resources may be congestible at higher levels Coarser Granularity: Resources may be congestible at higher levels
of granularity than bits or packets, for instance stateful of granularity than bits or packets, for instance stateful
firewalls are flow-congestible and call-servers are session- firewalls are flow-congestible and call-servers are session-
congestible. This memo focuses on congestion of connectionless congestible. This memo focuses on congestion of connectionless
resources, but the same principles may be applicable for resources, but the same principles may be applicable for
congestion notification protocols controlling per-flow and per- congestion notification protocols controlling per-flow and per-
session processing or state. session processing or state.
RED Terminology: In RED, whether to use packets or bytes when RED Terminology: In RED whether to use packets or bytes when
measuring queues is called respectively packet-mode queue measuring queues is called respectively "packet-mode queue
measurement or byte-mode queue measurement. And whether the measurement" or "byte-mode queue measurement". And whether the
probability of dropping a packet is independent or dependent on probability of dropping a particular packet is independent or
its byte-size is called respectively packet-mode drop or byte-mode dependent on its byte-size is called respectively "packet-mode
drop. The terms byte-mode and packet-mode should not be used drop" or "byte-mode drop". The terms byte-mode and packet-mode
without specifying whether they apply to queue measurement or to should not be used without specifying whether they apply to queue
drop. measurement or to drop.
2. Recommendations 1.2. Example Comparing Packet-Mode Drop and Byte-Mode Drop
A central question addressed by this document is whether to recommend
RED's packet-mode drop and to deprecate byte-mode drop. Table 1
compares how packet-mode and byte-mode drop affect two flows of
different size packets. For each it gives the expected number of
packets and of bits dropped in one second. Each example flow runs at
the same bit-rate of 48Mb/s, but one is broken up into small 60 byte
packets and the other into large 1500 byte packets.
To keep up the same bit-rate, in one second there are about 25 times
more small packets because they are 25 times smaller. As can be seen
from the table, the packet rate is 100,000 small packets versus 4,000
large packets per second (pps).
Parameter Formula Small packets Large packets
-------------------- -------------- ------------- -------------
Packet size s/8 60B 1,500B
Packet size s 480b 12,000b
Bit-rate x 48Mbps 48Mbps
Packet-rate u = x/s 100kpps 4kpps
Packet-mode Drop
Pkt loss probability p 0.1% 0.1%
Pkt loss-rate p*u 100pps 4pps
Bit loss-rate p*u*s 48kbps 48kbps
Byte-mode Drop MTU, M=12,000b
Pkt loss probability b = p*s/M 0.004% 0.1%
Pkt loss-rate b*u 4pps 4pps
Bit loss-rate b*u*s 1.92kbps 48kbps
Table 1: Example Comparing Packet-mode and Byte-mode Drop
For packet-mode drop, we illustrate the effect of a drop probability
of 0.1%, which the algorithm applies to all packets irrespective of
size. Because there are 25 times more small packets in one second,
it naturally drops 25 times more small packets, that is 100 small
packets but only 4 large packets. But if we count how many bits it
drops, there are 48,000 bits in 100 small packets and 48,000 bits in
4 large packets--the same number of bits of small packets as large.
The packet-mode drop algorithm drops any bit with the same
probability whether the bit is in a small or a large packet.
For byte-mode drop, again we use an example drop probability of 0.1%,
but only for maximum size packets (assuming the link MTU is 1,500B or
12,000b). The byte-mode algorithm reduces the drop probability of
smaller packets proportional to their size, making the probability
that it drops a small packet 25 times smaller at 0.004%. But there
are 25 times more small packets, so dropping them with 25 times lower
probability results in dropping the same number of packets: 4 drops
in both cases. The 4 small dropped packets contain 25 times less
bits than the 4 large dropped packets: 1,920 compared to 48,000.
The byte-mode drop algorithm drops any bit with a probability
proportionate to the size of the packet it is in.
2. Recommendations
2.1. Recommendation on Queue Measurement 2.1. Recommendation on Queue Measurement
Queue length is usually the most correct and simplest way to measure Queue length is usually the most correct and simplest way to measure
congestion of a resource. To avoid the pathological effects of drop congestion of a resource. To avoid the pathological effects of drop
tail, an AQM function can then be used to transform queue length into tail, an AQM function can then be used to transform queue length into
the probability of dropping or marking a packet (e.g. RED's the probability of dropping or marking a packet (e.g. RED's
piecewise linear function between thresholds). piecewise linear function between thresholds).
If the resource is bit-congestible, the implementation SHOULD measure If the resource is bit-congestible, the implementation SHOULD measure
the length of the queue in bytes. If the resource is packet- the length of the queue in bytes. If the resource is packet-
skipping to change at page 9, line 16 skipping to change at page 9, line 25
queue in packets. No other choice makes sense, because the number of queue in packets. No other choice makes sense, because the number of
packets waiting in the queue isn't relevant if the resource gets packets waiting in the queue isn't relevant if the resource gets
congested by bytes and vice versa. congested by bytes and vice versa.
Corollaries: Corollaries:
1. A RED implementation SHOULD use byte mode queue measurement for 1. A RED implementation SHOULD use byte mode queue measurement for
measuring the congestion of bit-congestible resources and packet measuring the congestion of bit-congestible resources and packet
mode queue measurement for packet-congestible resources. mode queue measurement for packet-congestible resources.
2. "An Admin SHOULD NOT be able to configure the way a queue 2. An implementation SHOULD NOT make it possible to configure the
measures itself, because wether a queue is bit-congestible or way a queue measures itself, because whether a queue is bit-
packet-congestible is a property of the resource." congestible or packet-congestible is an inherent property of the
queue.
The recommended approach in less straightforward scenarios, such as The recommended approach in less straightforward scenarios, such as
fixed size buffers, and resources without a queue, is discussed in fixed size buffers, and resources without a queue, is discussed in
Section 4.1. Section 4.1.
2.2. Recommendation on Notifying Congestion 2.2. Recommendation on Encoding Congestion Notification
When notifying congestion, a network device SHOULD treat all packets When encoding congestion notification (e.g. by drop, ECN & PCN), a
equally, regardless of their size. Therefore, the probability that network device SHOULD treat all packets equally, regardless of their
network equipment drops or marks a packet to notify congestion SHOULD size. In other words, the probability that network equipment drops
NOT depend on the size of the packet. For instance, to drop any bit or marks a particular packet to notify congestion SHOULD NOT depend
with probability 0.1% it is only necessary to drop every packet with on the size of the packet in question. As the example in Section 1.2
probability 0.1% without regard to the size of each packet. illustrates, to drop any bit with probability 0.1% it is only
necessary to drop every packet with probability 0.1% without regard
to the size of each packet.
This means that the Internet's congestion notification protocols This approach ensures the network layer offers sufficient congestion
(drop, ECN & PCN) SHOULD NOT take account of packet size when information for all known and future transport protocols and also
congestion is notified by network equipment. Allowance for packet ensures no perverse incentives are created that would encourage
size is only appropriate when the transport responds to congestion transports to use inappropriately small packet sizes.
(See Recommendation 2.3). This approach offers sufficient and
correct congestion information for all known and future transport
protocols and also ensures no perverse incentives are created that
would encourage transports to use inappropriately small packet sizes.
Corollaries: Corollaries:
1. AQM algorithms such as RED SHOULD NOT use byte-mode drop, which 1. AQM algorithms such as RED SHOULD NOT use byte-mode drop, which
deflates RED's drop probability for smaller packet sizes. RED's deflates RED's drop probability for smaller packet sizes. RED's
byte-mode drop has no enduring advantages. It is more complex, byte-mode drop has no enduring advantages. It is more complex,
it creates the perverse incentive to fragment segments into tiny it creates the perverse incentive to fragment segments into tiny
pieces and it reopens the vulnerability to floods of small- pieces and it reopens the vulnerability to floods of small-
packets that drop-tail queues suffered from and AQM was designed packets that drop-tail queues suffered from and AQM was designed
to remove. to remove.
2. If a vendor has implemented byte-mode drop, and an operator has 2. If a vendor has implemented byte-mode drop, and an operator has
turned it on, it is strongly RECOMMENDED that it SHOULD be turned turned it on, it is RECOMMENDED to turn it off. Note that RED as
off. Note that RED as a whole SHOULD NOT be turned off, as a whole SHOULD NOT be turned off, as without it, a drop tail
without it, a drop tail queue also biases against large packets. queue also biases against large packets. But note also that
But note also that turning off byte-mode drop may alter the turning off byte-mode drop may alter the relative performance of
relative performance of applications using different packet applications using different packet sizes, so it would be
sizes, so it would be advisable to establish the implications advisable to establish the implications before turning it off.
before turning it off.
NOTE WELL that RED's byte-mode queue drop is completely NOTE WELL that RED's byte-mode queue drop is completely
orthogonal to byte-mode queue measurement and should not be orthogonal to byte-mode queue measurement and should not be
confused with it. If a RED implementation has a byte-mode but confused with it. If a RED implementation has a byte-mode but
does not specify what sort of byte-mode, it is most probably does not specify what sort of byte-mode, it is most probably
byte-mode queue measurement, which is fine. However, if in byte-mode queue measurement, which is fine. However, if in
doubt, the vendor should be consulted. doubt, the vendor should be consulted.
The byte mode packet drop variant of RED was recommended in the past A survey (Appendix A) showed that there appears to be little, if any,
(see Section 4.2.1 for how thinking evolved). However, our survey of installed base of the byte-mode drop variant of RED. This suggests
84 vendors across the industry (Section 4.2.5) has found that none of that deprecating byte-mode drop will have little, if any, incremental
the 19% who responded have implemented byte mode drop in RED. Given
there appears to be little, if any, installed base it seems we can
deprecate byte-mode drop in RED with little, if any, incremental
deployment impact. deployment impact.
2.3. Recommendation on Responding to Congestion 2.3. Recommendation on Responding to Congestion
When a transport detects that a packet has been lost or congestion When a transport detects that a packet has been lost or congestion
marked, it SHOULD consider the strength of the congestion indication marked, it SHOULD consider the strength of the congestion indication
as proportionate to the size in octets of the missing or marked as proportionate to the size in octets (bytes) of the missing or
packet. marked packet.
In other words, when a packet indicates congestion (by being lost or In other words, when a packet indicates congestion (by being lost or
marked) it can be considered conceptually as if there is a congestion marked) it can be considered conceptually as if there is a congestion
indication on every octet of the packet, not just one indication per indication on every octet of the packet, not just one indication per
packet. packet.
Therefore, instead of network equipment biasing its congestion Therefore, the IETF transport area should continue its programme of;
notification in favour of small packets, the IETF transport area
should continue its programme of;
o updating host-based congestion control protocols to take account o updating host-based congestion control protocols to take account
of packet size of packet size
o making transports less sensitive to losing control packets like o making transports less sensitive to losing control packets like
SYNs and pure ACKs. SYNs and pure ACKs.
Corollaries: Corollaries:
1. If two TCPs with different packet sizes are required to run at 1. If two TCP flows with different packet sizes are required to run
equal bit rates under the same path conditions, this SHOULD be at equal bit rates under the same path conditions, this should be
done by altering TCP (Section 4.2.2), not network equipment, done by altering TCP (Section 4.2.2), not network equipment (the
which would otherwise affect other transports besides TCP. latter affects other transports besides TCP).
2. If it is desired to improve TCP performance by reducing the 2. If it is desired to improve TCP performance by reducing the
chance that a SYN or a pure ACK will be dropped, this should be chance that a SYN or a pure ACK will be dropped, this should be
done by modifying TCP (Section 4.2.3), not network equipment. done by modifying TCP (Section 4.2.3), not network equipment.
2.4. Recommendation on Handling Congestion Indications when Splitting 2.4. Recommendation on Handling Congestion Indications when Splitting
or Merging Packets or Merging Packets
Packets carrying congestion indications may be split or merged (e.g. Packets carrying congestion indications may be split or merged in
at a transcoder or during fragment reassembly). Splitting and some circumstances (e.g. at a RTCP transcoder or during IP fragment
merging only make sense in the context of ECN, not loss. reassembly). Splitting and merging only make sense in the context of
ECN, not loss.
The general rule to follow is that the number of octets in packets The general rule to follow is that the number of octets in packets
with congestion indications should be roughly the same before and with congestion indications SHOULD be equivalent before and after
after merging or splitting. This is based on the principle used merging or splitting. This is based on the principle used above;
above; that an indication of congestion on a packet can be considered that an indication of congestion on a packet can be considered as an
as an indication of congestion on each octet of the packet. indication of congestion on each octet of the packet.
The above rule is not phrased with the word "MUST" to allow the
following exception. There are cases where pre-existing protocols
were not designed to conserve congestion marked octets (e.g. IP
fragment reassembly [RFC3168] or loss statistics in RTCP receiver
reports [RFC3550] before ECN was added
[I-D.ietf-avtcore-ecn-for-rtp]). When any such protocol is updated,
it SHOULD comply with the above rule to conserved marked octets.
However, the rule may be relaxed if it would otherwise become too
complex to interoperate with pre-existing implementations of the
protocol.
One can think of a splitting or merging process as if all the One can think of a splitting or merging process as if all the
incoming congestion-marked octets increment a counter and all the incoming congestion-marked octets increment a counter and all the
outgoing marked octets decrement the same counter. In order to outgoing marked octets decrement the same counter. In order to
ensure that congestion indications remain timely, even the smallest ensure that congestion indications remain timely, even the smallest
positive remainder in the conceptual counter should trigger the next positive remainder in the conceptual counter should trigger the next
outgoing packet to be marked (causing the counter to go negative). outgoing packet to be marked (causing the counter to go negative).
3. Motivating Arguments 3. Motivating Arguments
In this section, we evaluate the topic of packet vs. byte based In this section, we justify the recommendations given in the previous
congestion notifications and motivate the recommendations given in section.
this document.
3.1. Scaling Congestion Control with Packet Size
There are two ways of interpreting a dropped or marked packet. It
can either be considered as a single loss event or as loss/marking of
the bytes in the packet.
Consider a bit-congestible link shared by many flows (bit-congestible
is the more common case, see Section 1.1), so that each busy period
tends to cause packets to be lost from different flows. Consider
further two sources that have the same data rate but break the load
into large packets in one application (A) and small packets in the
other (B). Of course, because the load is the same, there will be
proportionately more packets in the small packet flow (B).
If a congestion control scales with packet size it should respond in
the same way to the same congestion excursion, irrespective of the
size of the packets that the bytes causing congestion happen to be
broken down into.
A bit-congestible queue suffering a congestion excursion has to drop
or mark the same excess bytes whether they are in a few large packets
(A) or many small packets (B). So for the same congestion excursion,
the same amount of bytes have to be shed to get the load back to its
operating point. But, of course, for smaller packets (B) more
packets will have to be discarded to shed the same bytes.
If all the transports interpret each drop/mark as a single loss event
irrespective of the size of the packet dropped, those with smaller
packets (B) will respond more to the same congestion excursion. On
the other hand, if they respond proportionately less when smaller
packets are dropped/marked, overall they will be able to respond the
same to the same congestion excursion.
Therefore, for a congestion control to scale with packet size it
should respond to dropped or marked bytes (as TFRC-SP [RFC4828]
effectively does), instead of dropped or marked packets (as TCP
does).
3.2. Transport-Independent Network
TCP congestion control ensures that flows competing for the same
resource each maintain the same number of segments in flight,
irrespective of segment size. So under similar conditions, flows
with different segment sizes will get different bit rates.
Even though reducing the drop probability of small packets (e.g.
RED's byte-mode drop) helps ensure TCPs with different packet sizes
will achieve similar bit rates, we argue this correction should be
made to any future transport protocols based on TCP, not to the
network in order to fix one transport, no matter how prominent it is.
Effectively, favouring small packets is reverse engineering of
network equipment around one particular transport protocol (TCP),
contrary to the excellent advice in [RFC3426], which asks designers
to question "Why are you proposing a solution at this layer of the
protocol stack, rather than at another layer?"
RFC2309 refers to an email [pktByteEmail] for advice on how RED
should allow for different packet sizes. The email says the question
of whether a packet's own size should affect its drop probability
"depends on the dominant end-to-end congestion control mechanisms".
But we argue network equipment should not be specialised for whatever
transport is predominant. No matter how convenient it is, we SHOULD
NOT hack the network solely to allow for omissions from the design of
one transport protocol, even if it is as predominant as TCP.
3.3. Avoiding Perverse Incentives to (Ab)use Smaller Packets 3.1. Avoiding Perverse Incentives to (Ab)use Smaller Packets
Increasingly, it is being recognised that a protocol design must take Increasingly, it is being recognised that a protocol design must take
care not to cause unintended consequences by giving the parties in care not to cause unintended consequences by giving the parties in
the protocol exchange perverse incentives [Evol_cc][RFC3426]. Again, the protocol exchange perverse incentives [Evol_cc][RFC3426]. Given
imagine a scenario where the same bit rate of packets will contribute there are many good reasons why larger path max transmission units
(PMTUs) would help solve a number of scaling issues, we do not want
to create any bias against large packets that is greater than their
true cost.
Imagine a scenario where the same bit rate of packets will contribute
the same to bit-congestion of a link irrespective of whether it is the same to bit-congestion of a link irrespective of whether it is
sent as fewer larger packets or more smaller packets. A protocol sent as fewer larger packets or more smaller packets. A protocol
design that caused larger packets to be more likely to be dropped design that caused larger packets to be more likely to be dropped
than smaller ones would be dangerous in this case: than smaller ones would be dangerous in this case:
Malicious transports: A queue that gives an advantage to small Malicious transports: A queue that gives an advantage to small
packets can be used to amplify the force of a flooding attack. By packets can be used to amplify the force of a flooding attack. By
sending a flood of small packets, the attacker can get the queue sending a flood of small packets, the attacker can get the queue
to discard more traffic in large packets, allowing more attack to discard more traffic in large packets, allowing more attack
traffic to get through to cause further damage. Such a queue traffic to get through to cause further damage. Such a queue
allows attack traffic to have a disproportionately large effect on allows attack traffic to have a disproportionately large effect on
regular traffic without the attacker having to do much work. regular traffic without the attacker having to do much work.
Non-malicious transports: Even if a transport is not actually Non-malicious transports: Even if a transport designer is not
malicious, if it finds small packets go faster, over time it will actually malicious, if over time it is noticed that small packets
tend to act in its own interest and use them. Queues that give tend to go faster, designers will act in their own interest and
advantage to small packets create an evolutionary pressure for use smaller packets. Queues that give advantage to small packets
transports to send at the same bit-rate but break their data create an evolutionary pressure for transports to send at the same
stream down into tiny segments to reduce their drop rate. bit-rate but break their data stream down into tiny segments to
Encouraging a high volume of tiny packets might in turn reduce their drop rate. Encouraging a high volume of tiny packets
unnecessarily overload a completely unrelated part of the system, might in turn unnecessarily overload a completely unrelated part
perhaps more limited by header-processing than bandwidth. of the system, perhaps more limited by header-processing than
bandwidth.
Imagine two unresponsive flows arrive at a bit-congestible Imagine two unresponsive flows arrive at a bit-congestible
transmission link each with the same bit rate, say 1Mbps, but one transmission link each with the same bit rate, say 1Mbps, but one
consists of 1500B and the other 60B packets, which are 25x smaller. consists of 1500B and the other 60B packets, which are 25x smaller.
Consider a scenario where gentle RED [gentle_RED] is used, along with Consider a scenario where gentle RED [gentle_RED] is used, along with
the variant of RED we advise against, i.e. where the RED algorithm is the variant of RED we advise against, i.e. where the RED algorithm is
configured to adjust the drop probability of packets in proportion to configured to adjust the drop probability of packets in proportion to
each packet's size (byte mode packet drop). In this case, RED aims each packet's size (byte mode packet drop). In this case, RED aims
to drop 25x more of the larger packets than the smaller ones. Thus, to drop 25x more of the larger packets than the smaller ones. Thus,
for example if RED drops 25% of the larger packets, it will aim to for example if RED drops 25% of the larger packets, it will aim to
drop 1% of the smaller packets (but in practice it may drop more as drop 1% of the smaller packets (but in practice it may drop more as
congestion increases [RFC4828; S.B.4]). Even though both flows congestion increases [RFC4828; Appx B.4]). Even though both flows
arrive with the same bit rate, the bit rate the RED queue aims to arrive with the same bit rate, the bit rate the RED queue aims to
pass to the line will be 750Kbit for the flow of larger packet but pass to the line will be 750kbps for the flow of larger packets but
990Kbit for the smaller packets (but because of rate variation it 990kbps for the smaller packets (because of rate variations it will
will be less than this target). actually be a little less than this target).
Note that, although the byte-mode drop variant of RED amplifies small Note that, although the byte-mode drop variant of RED amplifies small
packet attacks, drop-tail queues amplify small packet attacks even packet attacks, drop-tail queues amplify small packet attacks even
more (see Security Considerations in Section 6). Wherever possible more (see Security Considerations in Section 6). Wherever possible
neither should be used. neither should be used.
3.4. Small != Control 3.2. Small != Control
It is tempting to drop small packets with lower probability to Dropping fewer control packets considerably improves performance. It
is tempting to drop small packets with lower probability in order to
improve performance, because many control packets are small (TCP SYNs improve performance, because many control packets are small (TCP SYNs
& ACKs, DNS queries & responses, SIP messages, HTTP GETs, etc) and & ACKs, DNS queries & responses, SIP messages, HTTP GETs, etc).
dropping fewer control packets considerably improves performance.
However, we must not give control packets preference purely by virtue However, we must not give control packets preference purely by virtue
of their smallness, otherwise it is too easy for any data source to of their smallness, otherwise it is too easy for any data source to
get the same preferential treatment simply by sending data in smaller get the same preferential treatment simply by sending data in smaller
packets. Again we should not create perverse incentives to favour packets. Again we should not create perverse incentives to favour
small packets rather than to favour control packets, which is what we small packets rather than to favour control packets, which is what we
intend. intend.
Just because many control packets are small does not mean all small Just because many control packets are small does not mean all small
packets are control packets. packets are control packets.
So again, rather than fix these problems in the network, we argue So, rather than fix these problems in the network, we argue that the
that the transport should be made more robust against losses of transport should be made more robust against losses of control
control packets (see 'Making Transports Robust against Control Packet packets (see 'Making Transports Robust against Control Packet Losses'
Losses' in Section 4.2.3). in Section 4.2.3).
3.5. Implementation Efficiency 3.3. Transport-Independent Network
Allowing for packet size at the transport rather than in the network TCP congestion control ensures that flows competing for the same
ensures that neither the network nor the transport needs to do a resource each maintain the same number of segments in flight,
multiply operation--multiplication by packet size is effectively irrespective of segment size. So under similar conditions, flows
achieved as a repeated add when the transport adds to its count of with different segment sizes will get different bit-rates.
marked bytes as each congestion event is fed to it. This isn't a
principled reason in itself, but it is a happy consequence of the
other principled reasons.
3.6. Why now? One motivation for the network biasing congestion notification by
packet size is to counter this effect and try to equalise the bit-
rates of flows with different packet sizes. However, in order to do
this, the queuing algorithm has to make assumptions about the
transport, which become embedded in the network. Specifically:
Now is a good time to discuss whether fairness between different o The queuing algorithm has to assume how aggressively the transport
sized packets would best be implemented in network equipment, or at will respond to congestion (see Section 4.2.4). If the network
the transport, for a number of reasons: assumes the transport responds as aggressively as TCP NewReno, it
will be wrong for Compound TCP and differently wrong for Cubic
TCP, etc. To achieve equal bit-rates, each transport then has to
guess what assumption the network made, and work out how to
replace this assumed aggressiveness with its own aggressiveness.
1. The IETF pre-congestion notification (PCN) working group is o Also, if the network biases congestion notification by packet size
standardising the external behaviour of a PCN congestion it has to assume a baseline packet size--all proposed algorithms
notification (AQM) algorithm [RFC5670]; use the local MTU. Then transports have to guess which link was
congested and what its local MTU was, in order to know how to
tailor their congestion response to that link.
2. [RFC2309] says RED may either take account of packet size or not Even though reducing the drop probability of small packets (e.g.
when dropping, but gives no recommendation between the two, RED's byte-mode drop) helps ensure TCP flows with different packet
referring instead to advice on the performance implications in an sizes will achieve similar bit rates, we argue this correction should
email [pktByteEmail], which recommends byte-mode drop. Further, be made to any future transport protocols based on TCP, not to the
just before RFC2309 was issued, an addendum was added to the network in order to fix one transport, no matter how predominant it
archived email that revisited the issue of packet vs. byte-mode is. Effectively, favouring small packets is reverse engineering of
drop in its last paragraph, making the recommendation less clear- network equipment around one particular transport protocol (TCP),
cut. RFC2309 is currently the only advice in the RFC series on contrary to the excellent advice in [RFC3426], which asks designers
packet size bias in AQM algorithms; to question "Why are you proposing a solution at this layer of the
protocol stack, rather than at another layer?"
3. The IRTF Internet Congestion Control Research Group (ICCRG) In contrast, if the network never takes account of packet size, the
recently took on the challenge of building consensus on what transport can be certain it will never need to guess any assumptions
common congestion control support should be required from network the network has made. And the network passes two pieces of
forwarding functions in future [RFC6077]. The wider Internet information to the transport that are sufficient in all cases: i)
community needs to discuss whether the complexity of adjusting congestion notification on the packet and ii) the size of the packet.
for packet size should be in the network or in transports; Both are available for the transport to combine (by taking account of
packet size when responding to congestion) or not. Appendix B checks
that these two pieces of information are sufficient for all relevant
scenarios.
4. Given there are many good reasons why larger path max When the network does not take account of packet size, it allows
transmission units (PMTUs) would help solve a number of scaling transport protocols to choose whether to take account of packet size
issues, we don't want to create any bias against large packets or not. However, if the network were to bias congestion notification
that is greater than their true cost; by packet size, transport protocols would have no choice; those that
did not take account of packet size themselves would unwittingly
become dependent on packet size, and those that already took account
of packet size would end up taking account of it twice.
5. The IETF audio/video transport (AVT) working group is 3.4. Scaling Congestion Control with Packet Size
standardising how the real-time protocol (RTP) should feedback
and respond to explicit congestion notification (ECN)
[I-D.ietf-avt-ecn-for-rtp].
6. The IETF has started to consider the question of fairness between Having so far justified only our recommendations for the network,
flows that use different packet sizes (e.g. in the small-packet this section focuses on the host. We construct a scaling argument to
variant of TCP-friendly rate control, TFRC-SP [RFC4828]). Given justify the recommendation that a host should respond to a dropped or
transports with different packet sizes, if we don't decide marked packet in proportion to its size, not just as a single
whether the network or the transport should allow for packet congestion event.
size, it will be hard if not impossible to design any transport
protocol so that its bit-rate relative to other transports meets The argument assumes that we have already sufficiently justified our
design guidelines [RFC5033] (Note however that, if the concern recommendation that the network should not take account of packet
were fairness between users, rather than between flows size.
[Rate_fair_Dis], relative rates between flows would have to come
under run-time control rather than being embedded in protocol Also, we assume bit-congestible links are the predominant source of
designs). congestion. As the Internet stands, it is hard if not impossible to
know whether congestion notification is from a bit-congestible or a
packet-congestible resource (see Appendix B.2) so we have to assume
the most prevalent case (see Section 1.1). If this assumption is
wrong, and particular congestion indications are actually due to
overload of packet-processing, there is no issue of safety at stake.
Any congestion control that triggers a multiplicative decrease in
response to a congestion indication will bring packet processing back
to its operating point just as quickly. The only issue at stake is
that the resource could be utilised more efficiently if packet-
congestion could be separately identified.
Imagine a bit-congestible link shared by many flows, so that each
busy period tends to cause packets to be lost from different flows.
Consider further two sources that have the same data rate but break
the load into large packets in one application (A) and small packets
in the other (B). Of course, because the load is the same, there
will be proportionately more packets in the small packet flow (B).
If a congestion control scales with packet size it should respond in
the same way to the same congestion notification, irrespective of the
size of the packets that the bytes causing congestion happen to be
broken down into.
A bit-congestible queue suffering congestion has to drop or mark the
same excess bytes whether they are in a few large packets (A) or many
small packets (B). So for the same amount of congestion overload,
the same amount of bytes has to be shed to get the load back to its
operating point. But, of course, for smaller packets (B) more
packets will have to be discarded to shed the same bytes.
If both the transports interpret each drop/mark as a single loss
event irrespective of the size of the packet dropped, the flow of
smaller packets (B) will respond more times to the same congestion.
On the other hand, if a transport responds proportionately less when
smaller packets are dropped/marked, overall it will be able to
respond the same to the same amount of congestion.
Therefore, for a congestion control to scale with packet size it
should respond to dropped or marked bytes (as TFRC-SP [RFC4828]
effectively does), instead of dropped or marked packets (as TCP
does).
For the avoidance of doubt, this is not a recommendation that TCP
should be changed so that it scales with packet size. It is a
recommendation that any future transport protocol proposal should
respond to dropped or marked bytes if it wishes to claim that it is
scalable.
3.5. Implementation Efficiency
Allowing for packet size at the transport rather than in the network
ensures that neither the network nor the transport needs to do a
multiply operation--multiplication by packet size is effectively
achieved as a repeated add when the transport adds to its count of
marked bytes as each congestion event is fed to it. This isn't a
principled reason in itself, but it is a happy consequence of the
other principled reasons.
4. A Survey and Critique of Past Advice 4. A Survey and Critique of Past Advice
This section is informative, not normative.
The original 1993 paper on RED [RED93] proposed two options for the The original 1993 paper on RED [RED93] proposed two options for the
RED active queue management algorithm: packet mode and byte mode. RED active queue management algorithm: packet mode and byte mode.
Packet mode measured the queue length in packets and dropped (or Packet mode measured the queue length in packets and dropped (or
marked) individual packets with a probability independent of their marked) individual packets with a probability independent of their
size. Byte mode measured the queue length in bytes and marked an size. Byte mode measured the queue length in bytes and marked an
individual packet with probability in proportion to its size individual packet with probability in proportion to its size
(relative to the maximum packet size). In the paper's outline of (relative to the maximum packet size). In the paper's outline of
further work, it was stated that no recommendation had been made on further work, it was stated that no recommendation had been made on
whether the queue size should be measured in bytes or packets, but whether the queue size should be measured in bytes or packets, but
noted that the difference could be significant. noted that the difference could be significant.
skipping to change at page 17, line 7 skipping to change at page 17, line 28
proportions of small packets, e.g. a DoS attack, and undersensitive proportions of small packets, e.g. a DoS attack, and undersensitive
to high proportions of large packets. However, there is no need to to high proportions of large packets. However, there is no need to
make allowances for the possibility of such legacy in future protocol make allowances for the possibility of such legacy in future protocol
design. This is safe because any undersensitivity during unusual design. This is safe because any undersensitivity during unusual
traffic mixes cannot lead to congestion collapse given the buffer traffic mixes cannot lead to congestion collapse given the buffer
will eventually revert to tail drop, discarding proportionately more will eventually revert to tail drop, discarding proportionately more
large packets. large packets.
4.1.1. Fixed Size Packet Buffers 4.1.1. Fixed Size Packet Buffers
Although the question of whether to measure queues in bytes or The question of whether to measure queues in bytes or packets seems
packets is fairly well understood these days, measuring congestion is to be well understood. However, measuring congestion is not
not straightforward when the resource is bit congestible but the straightforward when the resource is bit congestible but the queue is
queue is packet congestible or vice versa. This section outlines the packet congestible or vice versa. This section outlines the approach
approach to take. There is no controversy over what should be done, to take. There is no controversy over what should be done, you just
you just need to be expert in probability to work it out. And, even need to be expert in probability to work it out. And, even if you
if you know what should be done, it's not always easy to find a know what should be done, it's not always easy to find a practical
practical algorithm to implement it. algorithm to implement it.
Some, mostly older, queuing hardware sets aside fixed sized buffers Some, mostly older, queuing hardware sets aside fixed sized buffers
in which to store each packet in the queue. Also, with some in which to store each packet in the queue. Also, with some
hardware, any fixed sized buffers not completely filled by a packet hardware, any fixed sized buffers not completely filled by a packet
are padded when transmitted to the wire. If we imagine a theoretical are padded when transmitted to the wire. If we imagine a theoretical
forwarding system with both queuing and transmission in fixed, MTU- forwarding system with both queuing and transmission in fixed, MTU-
sized units, it should clearly be treated as packet-congestible, sized units, it should clearly be treated as packet-congestible,
because the queue length in packets would be a good model of because the queue length in packets would be a good model of
congestion of the lower layer link. congestion of the lower layer link.
skipping to change at page 18, line 23 skipping to change at page 18, line 44
simple rule for how to measure the length of queues of fixed buffers: simple rule for how to measure the length of queues of fixed buffers:
no matter how complicated the scheme is, ultimately any fixed buffer no matter how complicated the scheme is, ultimately any fixed buffer
system will need to measure its queue length in packets not bytes. system will need to measure its queue length in packets not bytes.
4.1.2. Congestion Measurement without a Queue 4.1.2. Congestion Measurement without a Queue
AQM algorithms are nearly always described assuming there is a queue AQM algorithms are nearly always described assuming there is a queue
for a congested resource and the algorithm can use the queue length for a congested resource and the algorithm can use the queue length
to determine the probability that it will drop or mark each packet. to determine the probability that it will drop or mark each packet.
But not all congested resources lead to queues. For instance, But not all congested resources lead to queues. For instance,
wireless spectrum is bit-congestible (for a given coding scheme), wireless spectrum is usually regarded as bit-congestible (for a given
because interference increases with the rate at which bits are coding scheme). But wireless link protocols do not always maintain a
transmitted. But wireless link protocols do not always maintain a
queue that depends on spectrum interference. Similarly, power queue that depends on spectrum interference. Similarly, power
limited resources are also usually bit-congestible if energy is limited resources are also usually bit-congestible if energy is
primarily required for transmission rather than header processing, primarily required for transmission rather than header processing,
but it is rare for a link protocol to build a queue as it approaches but it is rare for a link protocol to build a queue as it approaches
maximum power. maximum power.
Nonetheless, AQM algorithms do not require a queue in order to work. Nonetheless, AQM algorithms do not require a queue in order to work.
For instance spectrum congestion can be modelled by signal quality For instance spectrum congestion can be modelled by signal quality
using target bit-energy-to-noise-density ratio. And, to model radio using target bit-energy-to-noise-density ratio. And, to model radio
power exhaustion, transmission power levels can be measured and power exhaustion, transmission power levels can be measured and
compared to the maximum power available. [ECNFixedWireless] proposes compared to the maximum power available. [ECNFixedWireless] proposes
a practical and theoretically sound way to combine congestion a practical and theoretically sound way to combine congestion
notification for different bit-congestible resources at different notification for different bit-congestible resources at different
layers along an end to end path, whether wireless or wired, and layers along an end to end path, whether wireless or wired, and
whether with or without queues. whether with or without queues.
4.2. Congestion Notification Advice 4.2. Congestion Notification Advice
4.2.1. Network Bias when Encoding 4.2.1. Network Bias when Encoding
4.2.1.1. Advice on Packet Size Bias in RED
The previously mentioned email [pktByteEmail] referred to by The previously mentioned email [pktByteEmail] referred to by
[RFC2309] advised that most scarce resources in the Internet were [RFC2309] advised that most scarce resources in the Internet were
bit-congestible, which is still believed to be true (Section 1.1). bit-congestible, which is still believed to be true (Section 1.1).
But it went on to give advice we now disagree with. It said that But it went on to offer advice that is updated by this memo. It said
drop probability should depend on the size of the packet being that drop probability should depend on the size of the packet being
considered for drop if the resource is bit-congestible, but not if it considered for drop if the resource is bit-congestible, but not if it
is packet-congestible. The argument continued that if packet drops is packet-congestible. The argument continued that if packet drops
were inflated by packet size (byte-mode dropping), "a flow's fraction were inflated by packet size (byte-mode dropping), "a flow's fraction
of the packet drops is then a good indication of that flow's fraction of the packet drops is then a good indication of that flow's fraction
of the link bandwidth in bits per second". This was consistent with of the link bandwidth in bits per second". This was consistent with
a referenced policing mechanism being worked on at the time for a referenced policing mechanism being worked on at the time for
detecting unusually high bandwidth flows, eventually published in detecting unusually high bandwidth flows, eventually published in
1999 [pBox]. However, the problem could and should have been solved 1999 [pBox]. However, the problem could and should have been solved
by making the policing mechanism count the volume of bytes randomly by making the policing mechanism count the volume of bytes randomly
dropped, not the number of packets. dropped, not the number of packets.
skipping to change at page 19, line 33 skipping to change at page 20, line 7
In 2000, Cnodder et al [REDbyte] pointed out that there was an error In 2000, Cnodder et al [REDbyte] pointed out that there was an error
in the part of the original 1993 RED algorithm that aimed to in the part of the original 1993 RED algorithm that aimed to
distribute drops uniformly, because it didn't correctly take into distribute drops uniformly, because it didn't correctly take into
account the adjustment for packet size. They recommended an account the adjustment for packet size. They recommended an
algorithm called RED_4 to fix this. But they also recommended a algorithm called RED_4 to fix this. But they also recommended a
further change, RED_5, to adjust drop rate dependent on the square of further change, RED_5, to adjust drop rate dependent on the square of
relative packet size. This was indeed consistent with one implied relative packet size. This was indeed consistent with one implied
motivation behind RED's byte mode drop--that we should reverse motivation behind RED's byte mode drop--that we should reverse
engineer the network to improve the performance of dominant end-to- engineer the network to improve the performance of dominant end-to-
end congestion control mechanisms. But it is not consistent with the end congestion control mechanisms. This memo makes a different
present recommendations of Section 2. recommendations in Section 2.
By 2003, a further change had been made to the adjustment for packet By 2003, a further change had been made to the adjustment for packet
size, this time in the RED algorithm of the ns2 simulator. Instead size, this time in the RED algorithm of the ns2 simulator. Instead
of taking each packet's size relative to a `maximum packet size' it of taking each packet's size relative to a `maximum packet size' it
was taken relative to a `mean packet size', intended to be a static was taken relative to a `mean packet size', intended to be a static
value representative of the `typical' packet size on the link. We value representative of the `typical' packet size on the link. We
have not been able to find a justification in the literature for this have not been able to find a justification in the literature for this
change, however Eddy and Allman conducted experiments [REDbias] that change, however Eddy and Allman conducted experiments [REDbias] that
assessed how sensitive RED was to this parameter, amongst other assessed how sensitive RED was to this parameter, amongst other
things. No-one seems to have pointed out that this changed algorithm things. However, this changed algorithm can often lead to drop
can often lead to drop probabilities of greater than 1 (which should probabilities of greater than 1 (which gives a hint that there is
ring alarm bells hinting that there's a mistake in the theory probably a mistake in the theory somewhere).
somewhere).
On 10-Nov-2004, this variant of byte-mode packet drop was made the On 10-Nov-2004, this variant of byte-mode packet drop was made the
default in the ns2 simulator. None of the responses to our default in the ns2 simulator. It seems unlikely that byte-mode drop
admittedly limited survey of implementers (Section 4.2.5) found any has ever been implemented in production networks (Appendix A),
variant of byte-mode drop had been implemented. Therefore any therefore any conclusions based on ns2 simulations that use RED
conclusions based on ns2 simulations that use RED without disabling without disabling byte-mode drop are likely to behave very
byte-mode drop are likely to be highly questionable. differently from RED in production networks.
4.2.1.2. Packet Size Bias Regardless of RED
The byte-mode drop variant of RED is, of course, not the only The byte-mode drop variant of RED is, of course, not the only
possible bias towards small packets in queueing systems. We have possible bias towards small packets in queueing systems. We have
already mentioned that tail-drop queues naturally tend to lock-out already mentioned that tail-drop queues naturally tend to lock-out
large packets once they are full. But also queues with fixed sized large packets once they are full. But also queues with fixed sized
buffers reduce the probability that small packets will be dropped if buffers reduce the probability that small packets will be dropped if
(and only if) they allow small packets to borrow buffers from the (and only if) they allow small packets to borrow buffers from the
pools for larger packets. As was explained in Section 4.1.1 on fixed pools for larger packets. As was explained in Section 4.1.1 on fixed
size buffer carving, borrowing effectively makes the maximum queue size buffer carving, borrowing effectively makes the maximum queue
size for small packets greater than that for large packets, because size for small packets greater than that for large packets, because
skipping to change at page 20, line 42 skipping to change at page 21, line 17
lock-out large packets, purely because of the tail-drop aspect. So a lock-out large packets, purely because of the tail-drop aspect. So a
good AQM algorithm like RED with packet-mode drop should be used with good AQM algorithm like RED with packet-mode drop should be used with
fixed buffer memories where possible. If RED is too complicated to fixed buffer memories where possible. If RED is too complicated to
implement with multiple fixed buffer pools, the minimum necessary to implement with multiple fixed buffer pools, the minimum necessary to
prevent large packet lock-out is to ensure smaller packets never use prevent large packet lock-out is to ensure smaller packets never use
the last available buffer in any of the pools for larger packets. the last available buffer in any of the pools for larger packets.
4.2.2. Transport Bias when Decoding 4.2.2. Transport Bias when Decoding
The above proposals to alter the network equipment to bias towards The above proposals to alter the network equipment to bias towards
smaller packets have largely carried on outside the IETF process smaller packets have largely carried on outside the IETF process.
(unless one counts a reference in an informational RFC to an archived Whereas, within the IETF, there are many different proposals to alter
email!). Whereas, within the IETF, there are many different transport protocols to achieve the same goals, i.e. either to make
proposals to alter transport protocols to achieve the same goals, the flow bit-rate take account of packet size, or to protect control
i.e. either to make the flow bit-rate take account of packet size, or packets from loss. This memo argues that altering transport
to protect control packets from loss. This memo argues that altering protocols is the more principled approach.
transport protocols is the more principled approach.
A recently approved experimental RFC adapts its transport layer A recently approved experimental RFC adapts its transport layer
protocol to take account of packet sizes relative to typical TCP protocol to take account of packet sizes relative to typical TCP
packet sizes. This proposes a new small-packet variant of TCP- packet sizes. This proposes a new small-packet variant of TCP-
friendly rate control [RFC3448] called TFRC-SP [RFC4828]. friendly rate control [RFC5348] called TFRC-SP [RFC4828].
Essentially, it proposes a rate equation that inflates the flow rate Essentially, it proposes a rate equation that inflates the flow rate
by the ratio of a typical TCP segment size (1500B including TCP by the ratio of a typical TCP segment size (1500B including TCP
header) over the actual segment size [PktSizeEquCC]. (There are also header) over the actual segment size [PktSizeEquCC]. (There are also
other important differences of detail relative to TFRC, such as using other important differences of detail relative to TFRC, such as using
virtual packets [CCvarPktSize] to avoid responding to multiple losses virtual packets [CCvarPktSize] to avoid responding to multiple losses
per round trip and using a minimum inter-packet interval.) per round trip and using a minimum inter-packet interval.)
Section 4.5.1 of this TFRC-SP spec discusses the implications of Section 4.5.1 of this TFRC-SP spec discusses the implications of
operating in an environment where queues have been configured to drop operating in an environment where queues have been configured to drop
smaller packets with proportionately lower probability than larger smaller packets with proportionately lower probability than larger
skipping to change at page 21, line 38 skipping to change at page 22, line 12
conclusive, instead reporting simulations of many of the conclusive, instead reporting simulations of many of the
possibilities in order to assess performance but not recommending any possibilities in order to assess performance but not recommending any
particular course of action. particular course of action.
The paper originally proposing TFRC with virtual packets (VP-TFRC) The paper originally proposing TFRC with virtual packets (VP-TFRC)
[CCvarPktSize] proposed that there should perhaps be two variants to [CCvarPktSize] proposed that there should perhaps be two variants to
cater for the different variants of RED. However, as the TFRC-SP cater for the different variants of RED. However, as the TFRC-SP
authors point out, there is no way for a transport to know whether authors point out, there is no way for a transport to know whether
some queues on its path have deployed RED with byte-mode packet drop some queues on its path have deployed RED with byte-mode packet drop
(except if an exhaustive survey found that no-one has deployed it!-- (except if an exhaustive survey found that no-one has deployed it!--
see Section 4.2.4). Incidentally, VP-TFRC also proposed that byte- see Appendix A). Incidentally, VP-TFRC also proposed that byte-mode
mode RED dropping should really square the packet size compensation RED dropping should really square the packet-size compensation-factor
factor (like that of Cnodder's RED_5, but apparently unaware of it). (like that of Cnodder's RED_5, but apparently unaware of it).
Pre-congestion notification [RFC5670] is a proposal to use a virtual Pre-congestion notification [RFC5670] is an IETF technology to use a
queue for AQM marking for packets within one Diffserv class in order virtual queue for AQM marking for packets within one Diffserv class
to give early warning prior to any real queuing. The proposed PCN in order to give early warning prior to any real queuing. The PCN
marking algorithms have been designed not to take account of packet marking algorithms have been designed not to take account of packet
size when forwarding through queues. Instead the general principle size when forwarding through queues. Instead the general principle
has been to take account of the sizes of marked packets when has been to take account of the sizes of marked packets when
monitoring the fraction of marking at the edge of the network, as monitoring the fraction of marking at the edge of the network, as
recommended here. recommended here.
4.2.3. Making Transports Robust against Control Packet Losses 4.2.3. Making Transports Robust against Control Packet Losses
Recently, two RFCs have defined changes to TCP that make it more Recently, two RFCs have defined changes to TCP that make it more
robust against losing small control packets [RFC5562] [RFC5690]. In robust against losing small control packets [RFC5562] [RFC5690]. In
both cases they note that the case for these two TCP changes would be both cases they note that the case for these two TCP changes would be
weaker if RED were biased against dropping small packets. We argue weaker if RED were biased against dropping small packets. We argue
here that these two proposals are a safer and more principled way to here that these two proposals are a safer and more principled way to
achieve TCP performance improvements than reverse engineering RED to achieve TCP performance improvements than reverse engineering RED to
benefit TCP. benefit TCP.
Although no proposals exist as far as we know, it would also be Although there are no known proposals, it would also be possible and
possible and perfectly valid to make control packets robust against perfectly valid to make control packets robust against drop by
drop by explicitly requesting a lower drop probability using their explicitly requesting a lower drop probability using their Diffserv
Diffserv code point [RFC2474] to request a scheduling class with code point [RFC2474] to request a scheduling class with lower drop.
lower drop.
Although not brought to the IETF, a simple proposal from Wischik Although not brought to the IETF, a simple proposal from Wischik
[DupTCP] suggests that the first three packets of every TCP flow [DupTCP] suggests that the first three packets of every TCP flow
should be routinely duplicated after a short delay. It shows that should be routinely duplicated after a short delay. It shows that
this would greatly improve the chances of short flows completing this would greatly improve the chances of short flows completing
quickly, but it would hardly increase traffic levels on the Internet, quickly, but it would hardly increase traffic levels on the Internet,
because Internet bytes have always been concentrated in the large because Internet bytes have always been concentrated in the large
flows. It further shows that the performance of many typical flows. It further shows that the performance of many typical
applications depends on completion of long serial chains of short applications depends on completion of long serial chains of short
messages. It argues that, given most of the value people get from messages. It argues that, given most of the value people get from
skipping to change at page 22, line 45 skipping to change at page 23, line 16
+-----------+----------------+-----------------+--------------------+ +-----------+----------------+-----------------+--------------------+
| transport | RED_1 (packet | RED_4 (linear | RED_5 (square byte | | transport | RED_1 (packet | RED_4 (linear | RED_5 (square byte |
| cc | mode drop) | byte mode drop) | mode drop) | | cc | mode drop) | byte mode drop) | mode drop) |
+-----------+----------------+-----------------+--------------------+ +-----------+----------------+-----------------+--------------------+
| TCP or | s/sqrt(p) | sqrt(s/p) | 1/sqrt(p) | | TCP or | s/sqrt(p) | sqrt(s/p) | 1/sqrt(p) |
| TFRC | | | | | TFRC | | | |
| TFRC-SP | 1/sqrt(p) | 1/sqrt(sp) | 1/(s.sqrt(p)) | | TFRC-SP | 1/sqrt(p) | 1/sqrt(sp) | 1/(s.sqrt(p)) |
+-----------+----------------+-----------------+--------------------+ +-----------+----------------+-----------------+--------------------+
Table 1: Dependence of flow bit-rate per RTT on packet size s and Table 2: Dependence of flow bit-rate per RTT on packet size, s, and
drop rate p when network and/or transport bias towards small packets drop probability, p, when network and/or transport bias towards small
to varying degrees packets to varying degrees
Table 1 aims to summarise the potential effects of all the advice Table 2 aims to summarise the potential effects of all the advice
from different sources. Each column shows a different possible AQM from different sources. Each column shows a different possible AQM
behaviour in different queues in the network, using the terminology behaviour in different queues in the network, using the terminology
of Cnodder et al outlined earlier (RED_1 is basic RED with packet- of Cnodder et al outlined earlier (RED_1 is basic RED with packet-
mode drop). Each row shows a different transport behaviour: TCP mode drop). Each row shows a different transport behaviour: TCP
[RFC5681] and TFRC [RFC3448] on the top row with TFRC-SP [RFC4828] [RFC5681] and TFRC [RFC5348] on the top row with TFRC-SP [RFC4828]
below. below. Each cell shows how the bits per round trip of a flow depends
on packet size, s, and drop probability, p. In order to declutter
the formulae to focus on packet-size dependence they are all given
per round trip, which removes any RTT term.
Let us assume that the goal is for the bit-rate of a flow to be Let us assume that the goal is for the bit-rate of a flow to be
independent of packet size. Suppressing all inessential details, the independent of packet size. Suppressing all inessential details, the
table shows that this should either be achievable by not altering the table shows that this should either be achievable by not altering the
TCP transport in a RED_5 network, or using the small packet TFRC-SP TCP transport in a RED_5 network, or using the small packet TFRC-SP
transport (or similar) in a network without any byte-mode dropping transport (or similar) in a network without any byte-mode dropping
RED (top right and bottom left). Top left is the `do nothing' RED (top right and bottom left). Top left is the `do nothing'
scenario, while bottom right is the `do-both' scenario in which bit- scenario, while bottom right is the `do-both' scenario in which bit-
rate would become far too biased towards small packets. Of course, rate would become far too biased towards small packets. Of course,
if any form of byte-mode dropping RED has been deployed on a subset if any form of byte-mode dropping RED has been deployed on a subset
of queues that congest, each path through the network will present a of queues that congest, each path through the network will present a
different hybrid scenario to its transport. different hybrid scenario to its transport.
Whatever, we can see that the linear byte-mode drop column in the Whatever, we can see that the linear byte-mode drop column in the
middle considerably complicates the Internet. It's a half-way house middle would considerably complicate the Internet. It's a half-way
that doesn't bias enough towards small packets even if one believes house that doesn't bias enough towards small packets even if one
the network should be doing the biasing. Section 2 recommends that believes the network should be doing the biasing. Section 2
_all_ bias in network equipment towards small packets should be recommends that _all_ bias in network equipment towards small packets
turned off--if indeed any equipment vendors have implemented it-- should be turned off--if indeed any equipment vendors have
leaving packet size bias solely as the preserve of the transport implemented it--leaving packet-size bias solely as the preserve of
layer (solely the leftmost, packet-mode drop column). the transport layer (solely the leftmost, packet-mode drop column).
4.2.5. RED Implementation Status
A survey has been conducted of 84 vendors to assess how widely drop
probability based on packet size has been implemented in RED. Prior
to the survey, an individual approach to Cisco received confirmation
that, having checked the code-base for each of the product ranges,
Cisco has not implemented any discrimination based on packet size in
any AQM algorithm in any of its products. Also an individual
approach to Alcatel-Lucent drew a confirmation that it was very
likely that none of their products contained RED code that
implemented any packet-size bias.
Turning to our more formal survey (Table 2), about 19% of those
surveyed have replied so far, giving a sample size of 16. Although
we do not have permission to identify the respondents, we can say
that those that have responded include most of the larger vendors,
covering a large fraction of the market. They range across the large
network equipment vendors at L3 & L2, firewall vendors, wireless
equipment vendors, as well as large software businesses with a small
selection of networking products. So far, all those who have
responded have confirmed that they have not implemented the variant
of RED with drop dependent on packet size (2 were fairly sure they
had not but needed to check more thoroughly). We have established
that Linux does not implement RED with packet size drop bias,
although we have not investigated a wider range of open source code.
+-------------------------------+----------------+-----------------+
| Response | No. of vendors | %age of vendors |
+-------------------------------+----------------+-----------------+
| Not implemented | 14 | 17% |
| Not implemented (probably) | 2 | 2% |
| Implemented | 0 | 0% |
| No response | 68 | 81% |
| Total companies/orgs surveyed | 84 | 100% |
+-------------------------------+----------------+-----------------+
Table 2: Vendor Survey on byte-mode drop variant of RED (lower drop
probability for small packets)
Where reasons have been given, the extra complexity of packet bias
code has been most prevalent, though one vendor had a more principled
reason for avoiding it--similar to the argument of this document.
Finally, we repeat that RED's byte mode drop SHOULD be disabled, but
active queue management such as RED SHOULD be enabled wherever
possible if we are to eradicate bias towards small packets--without
any AQM at all, tail-drop tends to lock-out large packets very
effectively.
Our survey was of vendor implementations, so we cannot be certain
about operator deployment. But we believe many queues in the
Internet are still tail-drop. The company of one of the co-authors
(BT) has widely deployed RED, but many tail-drop queues are there are
bound to still exist, particularly in access network equipment and on
middleboxes like firewalls, where RED is not always available.
Routers using a memory architecture based on fixed size buffers with In practice it seems that no deliberate bias towards small packets
borrowing may also still be prevalent in the Internet. As explained has been implemented for production networks. Of the 19% of vendors
in Section 4.2.1, these also provide a marginal (but legitimate) bias who responded to a survey of 84 equipment vendors, none had
towards small packets. So even though RED byte-mode drop is not implemented byte-mode drop in RED (see Appendix A for details).
prevalent, it is likely there is still some bias towards small
packets in the Internet due to tail drop and fixed buffer borrowing.
5. Outstanding Issues and Next Steps 5. Outstanding Issues and Next Steps
5.1. Bit-congestible World 5.1. Bit-congestible Network
For a connectionless network with nearly all resources being bit- For a connectionless network with nearly all resources being bit-
congestible we believe the recommended position is now unarguably congestible the recommended position is clear--that the network
clear--that the network should not make allowance for packet sizes should not make allowance for packet sizes and the transport should.
and the transport should. This leaves two outstanding issues: This leaves two outstanding issues:
o How to handle any legacy of AQM with byte-mode drop already o How to handle any legacy of AQM with byte-mode drop already
deployed; deployed;
o The need to start a programme to update transport congestion o The need to start a programme to update transport congestion
control protocol standards to take account of packet size. control protocol standards to take account of packet size.
The sample of returns from our vendor survey Section 4.2.4 suggest A survey of equipment vendors (Section 4.2.4) found no evidence that
that byte-mode packet drop seems not to be implemented at all let byte-mode packet drop had been implemented, so deployment will be
alone deployed, or if it is, it is likely to be very sparse. sparse at best. A migration strategy is not really needed to remove
Therefore, we do not really need a migration strategy from all but an algorithm that may not even be deployed.
nothing to nothing.
A programme of standards updates to take account of packet size in
transport congestion control protocols has started with TFRC-SP
[RFC4828], while weighted TCPs implemented in the research community
[WindowPropFair] could form the basis of a future change to TCP
congestion control [RFC5681] itself.
5.2. Bit- & Packet-congestible World
Nonetheless, the position is much less clear-cut if the Internet
becomes populated by a more even mix of both packet-congestible and
bit-congestible resources. If we believe we should allow for this
possibility in the future, this space contains a truly open research
issue.
We develop the concept of an idealised congestion notification
protocol that supports both bit-congestible and packet-congestible
resources in Appendix A. This congestion notification requires at
least two flags for congestion of bit-congestible and packet-
congestible resources. This hides a fundamental problem--much more
fundamental than whether we can magically create header space for yet
another ECN flag in IPv4, or whether it would work while being
deployed incrementally. Distinguishing drop from delivery naturally
provides just one congestion flag--it is hard to drop a packet in two
ways that are distinguishable remotely. This is a similar problem to
that of distinguishing wireless transmission losses from congestive
losses.
This problem would not be solved even if ECN were universally A programme of experimental updates to take account of packet size in
deployed. A congestion notification protocol must survive a transport congestion control protocols has already started with
transition from low levels of congestion to high. Marking two states TFRC-SP [RFC4828].
is feasible with explicit marking, but much harder if packets are
dropped. Also, it will not always be cost-effective to implement AQM
at every low level resource, so drop will often have to suffice.
We should also note that, strictly, packet-congestible resources are 5.2. Bit- & Packet-congestible Network
actually cycle-congestible because load also depends on the
complexity of each look-up and whether the pattern of arrivals is
amenable to caching or not. Further, this reminds us that any
solution must not require a forwarding engine to use excessive
processor cycles in order to decide how to say it has no spare
processor cycles.
Recently, the dual resource queue (DRQ) proposal [DRQ] has been made The position is much less clear-cut if the Internet becomes populated
on the premise that, as network processors become more cost by a more even mix of both packet-congestible and bit-congestible
effective, per packet operations will become more complex resources (see Appendix B.2). This problem is not pressing, because
(irrespective of whether more function in the network is desirable). most Internet resources are designed to be bit-congestible before
Consequently the premise is that CPU congestion will become more packet processing starts to congest (see Section 1.1).
common. DRQ is a proposed modification to the RED algorithm that
folds both bit congestion and packet congestion into one signal
(either loss or ECN).
The problem of signalling packet processing congestion is not The IRTF Internet congestion control research group (ICCRG) has set
pressing, as most Internet resources are designed to be bit- itself the task of reaching consensus on generic forwarding
congestible before packet processing starts to congest (see mechanisms that are necessary and sufficient to support the
Section 1.1). However, the IRTF Internet congestion control research Internet's future congestion control requirements (the first
group (ICCRG) has set itself the task of reaching consensus on challenge in [RFC6077]). Therefore, we defer the question of whether
generic forwarding mechanisms that are necessary and sufficient to packet congestion might become common and what to do if it does to
support the Internet's future congestion control requirements (the the IRTF (the 'Small Packets' challenge in [RFC6077]).
first challenge in [RFC6077]). Therefore, rather than not giving
this problem any thought at all, just because it is hard and
currently hypothetical, we defer the question of whether packet
congestion might become common and what to do if it does to the IRTF
(the 'Small Packets' challenge in [RFC6077]).
6. Security Considerations 6. Security Considerations
This draft recommends that queues do not bias drop probability This memo recommends that queues do not bias drop probability towards
towards small packets as this creates a perverse incentive for small packets as this creates a perverse incentive for transports to
transports to break down their flows into tiny segments. One of the break down their flows into tiny segments. One of the benefits of
benefits of implementing AQM was meant to be to remove this perverse implementing AQM was meant to be to remove this perverse incentive
incentive that drop-tail queues gave to small packets. Of course, if that drop-tail queues gave to small packets.
transports really want to make the greatest gains, they don't have to
respond to congestion anyway. But we don't want applications that
are trying to behave to discover that they can go faster by using
smaller packets.
In practice, transports cannot all be trusted to respond to In practice, transports cannot all be trusted to respond to
congestion. So another reason for recommending that queues do not congestion. So another reason for recommending that queues do not
bias drop probability towards small packets is to avoid the bias drop probability towards small packets is to avoid the
vulnerability to small packet DDoS attacks that would otherwise vulnerability to small packet DDoS attacks that would otherwise
result. One of the benefits of implementing AQM was meant to be to result. One of the benefits of implementing AQM was meant to be to
remove drop-tail's DoS vulnerability to small packets, so we remove drop-tail's DoS vulnerability to small packets, so we
shouldn't add it back again. shouldn't add it back again.
If most queues implemented AQM with byte-mode drop, the resulting If most queues implemented AQM with byte-mode drop, the resulting
network would amplify the potency of a small packet DDoS attack. At network would amplify the potency of a small packet DDoS attack. At
the first queue the stream of packets would push aside a greater the first queue the stream of packets would push aside a greater
proportion of large packets, so more of the small packets would proportion of large packets, so more of the small packets would
survive to attack the next queue. Thus a flood of small packets survive to attack the next queue. Thus a flood of small packets
would continue on towards the destination, pushing regular traffic would continue on towards the destination, pushing regular traffic
with large packets out of the way in one queue after the next, but with large packets out of the way in one queue after the next, but
suffering much less drop itself. suffering much less drop itself.
Appendix B explains why the ability of networks to police the Appendix C explains why the ability of networks to police the
response of _any_ transport to congestion depends on bit-congestible response of _any_ transport to congestion depends on bit-congestible
network resources only doing packet-mode not byte-mode drop. In network resources only doing packet-mode not byte-mode drop. In
summary, it says that making drop probability depend on the size of summary, it says that making drop probability depend on the size of
the packets that bits happen to be divided into simply encourages the the packets that bits happen to be divided into simply encourages the
bits to be divided into smaller packets. Byte-mode drop would bits to be divided into smaller packets. Byte-mode drop would
therefore irreversibly complicate any attempt to fix the Internet's therefore irreversibly complicate any attempt to fix the Internet's
incentive structures. incentive structures.
7. Conclusions 7. Conclusions
This memo strongly recommends that the size of an individual packet This memo identifies the three distinct stages of the congestion
that is dropped or marked should only be taken into account when a notification process where implementations need to decide whether to
transport reads this as a congestion indication, not when network take packet size into account. The recommendation of this memo is
equipment writes it. The memo therefore strongly deprecates using different in each case:
RED's byte-mode of packet drop in network equipment.
Whether network equipment should measure the length of a queue by o When network equipment measures the length of a queue, whether it
counting bytes or counting packets is a different question to whether counts in bytes or packets depends on whether the network resource
it should take into account the size of each packet being dropped or is congested respectively by bytes or by packets.
marked. The answer depends on whether the network resource is
congested respectively by bytes or by packets. This means that RED's o When network equipment decides whether to drop (or mark) a packet,
byte-mode queue measurement will often be appropriate even though it is recommended that the size of the particular packet should
byte-mode drop is strongly deprecated. not be taken into account
o However, when a transport algorithm responds to a dropped or
marked packet, the size of the rate reduction should be
proportionate to the size of the packet.
In summary, the answers are 'it depends', 'no' and 'yes' respectively
This means that RED's byte-mode queue measurement will often be
appropriate although byte-mode drop is strongly deprecated.
At the transport layer the IETF should continue updating congestion At the transport layer the IETF should continue updating congestion
control protocols to take account of the size of each packet that control protocols to take account of the size of each packet that
indicates congestion. Also the IETF should continue to make indicates congestion. Also the IETF should continue to make
transports less sensitive to losing control packets like SYNs, pure protocols less sensitive to losing control packets like SYNs, pure
ACKs and DNS exchanges. Although many control packets happen to be ACKs and DNS exchanges. Although many control packets happen to be
small, the alternative of network equipment favouring all small small, the alternative of network equipment favouring all small
packets would be dangerous. That would create perverse incentives to packets would be dangerous. That would create perverse incentives to
split data transfers into smaller packets. split data transfers into smaller packets.
The memo develops these recommendations from principled arguments The memo develops these recommendations from principled arguments
concerning scaling, layering, incentives, inherent efficiency, concerning scaling, layering, incentives, inherent efficiency,
security and policability. But it also addresses practical issues security and policeability. But it also addresses practical issues
such as specific buffer architectures and incremental deployment. such as specific buffer architectures and incremental deployment.
Indeed a limited survey of RED implementations is included, which Indeed a limited survey of RED implementations is discussed, which
shows there appears to be little, if any, installed base of RED's shows there appears to be little, if any, installed base of RED's
byte-mode drop. Therefore it can be deprecated with little, if any, byte-mode drop. Therefore it can be deprecated with little, if any,
incremental deployment complications. incremental deployment complications.
The recommendations have been developed on the well-founded basis The recommendations have been developed on the well-founded basis
that most Internet resources are bit-congestible not packet- that most Internet resources are bit-congestible not packet-
congestible. We need to know the likelihood that this assumption congestible. We need to know the likelihood that this assumption
will prevail longer term and, if it might not, what protocol changes will prevail longer term and, if it might not, what protocol changes
will be needed to cater for a mix of the two. These questions have will be needed to cater for a mix of the two. This problem is
been delegated to the IRTF. deferred to the IRTF Internet Congestion Control Research Group
(ICCRG).
8. Acknowledgements 8. Acknowledgements
Thank you to Sally Floyd, who gave extensive and useful review Thank you to Sally Floyd, who gave extensive and useful review
comments. Also thanks for the reviews from Philip Eardley, Toby comments. Also thanks for the reviews from Philip Eardley, David
Moncaster, Arnaud Jacquet and Mirja Kuehlewind as well as helpful Black, Fred Baker, Toby Moncaster, Arnaud Jacquet and Mirja
explanations of different hardware approaches from Larry Dunn and Kuehlewind as well as helpful explanations of different hardware
Fred Baker. We are grateful to Bruce Davie and his colleagues for approaches from Larry Dunn and Fred Baker. We are grateful to Bruce
providing a timely and efficient survey of RED implementation in Davie and his colleagues for providing a timely and efficient survey
Cisco's product range. Also grateful thanks to Toby Moncaster, Will of RED implementation in Cisco's product range. Also grateful thanks
Dormann, John Regnault, Simon Carter and Stefaan De Cnodder who to Toby Moncaster, Will Dormann, John Regnault, Simon Carter and
further helped survey the current status of RED implementation and Stefaan De Cnodder who further helped survey the current status of
deployment and, finally, thanks to the anonymous individuals who RED implementation and deployment and, finally, thanks to the
responded. anonymous individuals who responded.
Bob Briscoe and Jukka Manner are partly funded by Trilogy, a research Bob Briscoe and Jukka Manner are partly funded by Trilogy, a research
project (ICT- 216372) supported by the European Community under its project (ICT- 216372) supported by the European Community under its
Seventh Framework Programme. The views expressed here are those of Seventh Framework Programme. The views expressed here are those of
the authors only. the authors only.
9. Comments Solicited 9. Comments Solicited
Comments and questions are encouraged and very welcome. They can be Comments and questions are encouraged and very welcome. They can be
addressed to the IETF Transport Area working group mailing list addressed to the IETF Transport Area working group mailing list
skipping to change at page 29, line 19 skipping to change at page 27, line 39
[RFC3168] Ramakrishnan, K., Floyd, S., and D. [RFC3168] Ramakrishnan, K., Floyd, S., and D.
Black, "The Addition of Explicit Black, "The Addition of Explicit
Congestion Notification (ECN) to IP", Congestion Notification (ECN) to IP",
RFC 3168, September 2001. RFC 3168, September 2001.
[RFC3426] Floyd, S., "General Architectural and [RFC3426] Floyd, S., "General Architectural and
Policy Considerations", RFC 3426, Policy Considerations", RFC 3426,
November 2002. November 2002.
[RFC5033] Floyd, S. and M. Allman, "Specifying
New Congestion Control Algorithms",
BCP 133, RFC 5033, August 2007.
10.2. Informative References 10.2. Informative References
[CCvarPktSize] Widmer, J., Boutremans, C., and J-Y. [CCvarPktSize] Widmer, J., Boutremans, C., and J-Y.
Le Boudec, "Congestion Control for Le Boudec, "Congestion Control for
Flows with Variable Packet Size", ACM Flows with Variable Packet Size", ACM
CCR 34(2) 137--151, 2004, <http:// CCR 34(2) 137--151, 2004, <http://
doi.acm.org/10.1145/997150.997162>. doi.acm.org/10.1145/997150.997162>.
[CHOKe_Var_Pkt] Psounis, K., Pan, R., and B.
Prabhaker, "Approximate Fair Dropping
for Variable Length Packets", IEEE
Micro 21(1):48--56, January-
February 2001, <http://
www.stanford.edu/~balaji/papers/
01approximatefair.pdf}>.
[DRQ] Shin, M., Chong, S., and I. Rhee, [DRQ] Shin, M., Chong, S., and I. Rhee,
"Dual-Resource TCP/AQM for "Dual-Resource TCP/AQM for
Processing-Constrained Networks", Processing-Constrained Networks",
IEEE/ACM Transactions on IEEE/ACM Transactions on
Networking Vol 16, issue 2, Networking Vol 16, issue 2,
April 2008, <http://dx.doi.org/ April 2008, <http://dx.doi.org/
10.1109/TNET.2007.900415>. 10.1109/TNET.2007.900415>.
[DupTCP] Wischik, D., "Short messages", Royal [DupTCP] Wischik, D., "Short messages", Royal
Society workshop on networks: Society workshop on networks:
skipping to change at page 30, line 13 skipping to change at page 28, line 37
resource_control_elastic_cdma.html>. resource_control_elastic_cdma.html>.
[Evol_cc] Gibbens, R. and F. Kelly, "Resource [Evol_cc] Gibbens, R. and F. Kelly, "Resource
pricing and the evolution of pricing and the evolution of
congestion control", congestion control",
Automatica 35(12)1969--1985, Automatica 35(12)1969--1985,
December 1999, <http:// December 1999, <http://
www.statslab.cam.ac.uk/~frank/ www.statslab.cam.ac.uk/~frank/
evol.html>. evol.html>.
[I-D.ietf-avt-ecn-for-rtp] Westerlund, M., Johansson, I., [I-D.ietf-avtcore-ecn-for-rtp] Westerlund, M., Johansson, I.,
Perkins, C., and K. Carlberg, Perkins, C., O'Hanlon, P., and K.
"Explicit Congestion Notification Carlberg, "Explicit Congestion
(ECN) for RTP over UDP", Notification (ECN) for RTP over UDP",
draft-ietf-avt-ecn-for-rtp-03 (work draft-ietf-avtcore-ecn-for-rtp-04
in progress), October 2010. (work in progress), July 2011.
[I-D.ietf-conex-concepts-uses] Briscoe, B., Woundy, R., Moncaster, [I-D.ietf-conex-concepts-uses] Briscoe, B., Woundy, R., and A.
T., and J. Leslie, "ConEx Concepts Cooper, "ConEx Concepts and Use
and Use Cases", Cases",
draft-ietf-conex-concepts-uses-00 draft-ietf-conex-concepts-uses-03
(work in progress), November 2010. (work in progress), October 2011.
[IOSArch] Bollapragada, V., White, R., and C. [IOSArch] Bollapragada, V., White, R., and C.
Murphy, "Inside Cisco IOS Software Murphy, "Inside Cisco IOS Software
Architecture", Cisco Press: CCIE Architecture", Cisco Press: CCIE
Professional Development ISBN13: 978- Professional Development ISBN13: 978-
1-57870-181-0, July 2000. 1-57870-181-0, July 2000.
[MulTCP] Crowcroft, J. and Ph. Oechslin,
"Differentiated End to End Internet
Services using a Weighted
Proportional Fair Sharing TCP",
CCR 28(3) 53--69, July 1998, <http://
www.cs.ucl.ac.uk/staff/J.Crowcroft/
hipparch/pricing.html>.
[PktSizeEquCC] Vasallo, P., "Variable Packet Size [PktSizeEquCC] Vasallo, P., "Variable Packet Size
Equation-Based Congestion Control", Equation-Based Congestion Control",
ICSI Technical Report tr-00-008, ICSI Technical Report tr-00-008,
2000, <http://http.icsi.berkeley.edu/ 2000, <http://http.icsi.berkeley.edu/
ftp/global/pub/techreports/2000/ ftp/global/pub/techreports/2000/
tr-00-008.pdf>. tr-00-008.pdf>.
[RED93] Floyd, S. and V. Jacobson, "Random [RED93] Floyd, S. and V. Jacobson, "Random
Early Detection (RED) gateways for Early Detection (RED) gateways for
Congestion Avoidance", IEEE/ACM Congestion Avoidance", IEEE/ACM
skipping to change at page 31, line 26 skipping to change at page 29, line 43
Communications (ISCC) 793--799, Communications (ISCC) 793--799,
July 2000, <http://www.icir.org/ July 2000, <http://www.icir.org/
floyd/red/Elloumi99.pdf>. floyd/red/Elloumi99.pdf>.
[RFC2474] Nichols, K., Blake, S., Baker, F., [RFC2474] Nichols, K., Blake, S., Baker, F.,
and D. Black, "Definition of the and D. Black, "Definition of the
Differentiated Services Field (DS Differentiated Services Field (DS
Field) in the IPv4 and IPv6 Headers", Field) in the IPv4 and IPv6 Headers",
RFC 2474, December 1998. RFC 2474, December 1998.
[RFC3448] Handley, M., Floyd, S., Padhye, J., [RFC3550] Schulzrinne, H., Casner, S.,
and J. Widmer, "TCP Friendly Rate Frederick, R., and V. Jacobson, "RTP:
Control (TFRC): Protocol A Transport Protocol for Real-Time
Specification", RFC 3448, Applications", STD 64, RFC 3550,
January 2003. July 2003.
[RFC3714] Floyd, S. and J. Kempf, "IAB Concerns [RFC3714] Floyd, S. and J. Kempf, "IAB Concerns
Regarding Congestion Control for Regarding Congestion Control for
Voice Traffic in the Internet", Voice Traffic in the Internet",
RFC 3714, March 2004. RFC 3714, March 2004.
[RFC4828] Floyd, S. and E. Kohler, "TCP [RFC4828] Floyd, S. and E. Kohler, "TCP
Friendly Rate Control (TFRC): The Friendly Rate Control (TFRC): The
Small-Packet (SP) Variant", RFC 4828, Small-Packet (SP) Variant", RFC 4828,
April 2007. April 2007.
[RFC5348] Floyd, S., Handley, M., Padhye, J.,
and J. Widmer, "TCP Friendly Rate
Control (TFRC): Protocol
Specification", RFC 5348,
September 2008.
[RFC5562] Kuzmanovic, A., Mondal, A., Floyd, [RFC5562] Kuzmanovic, A., Mondal, A., Floyd,
S., and K. Ramakrishnan, "Adding S., and K. Ramakrishnan, "Adding
Explicit Congestion Notification Explicit Congestion Notification
(ECN) Capability to TCP's SYN/ACK (ECN) Capability to TCP's SYN/ACK
Packets", RFC 5562, June 2009. Packets", RFC 5562, June 2009.
[RFC5670] Eardley, P., "Metering and Marking [RFC5670] Eardley, P., "Metering and Marking
Behaviour of PCN-Nodes", RFC 5670, Behaviour of PCN-Nodes", RFC 5670,
November 2009. November 2009.
skipping to change at page 32, line 25 skipping to change at page 30, line 47
M., and B. Briscoe, "Open Research M., and B. Briscoe, "Open Research
Issues in Internet Congestion Issues in Internet Congestion
Control", RFC 6077, February 2011. Control", RFC 6077, February 2011.
[Rate_fair_Dis] Briscoe, B., "Flow Rate Fairness: [Rate_fair_Dis] Briscoe, B., "Flow Rate Fairness:
Dismantling a Religion", ACM Dismantling a Religion", ACM
CCR 37(2)63--74, April 2007, <http:// CCR 37(2)63--74, April 2007, <http://
portal.acm.org/ portal.acm.org/
citation.cfm?id=1232926>. citation.cfm?id=1232926>.
[WindowPropFair] Siris, V., "Service Differentiation
and Performance of Weighted Window-
Based Congestion Control and Packet
Marking Algorithms in ECN Networks",
Computer Communications 26(4) 314--
326, 2002, <http://www.ics.forth.gr/
netgroup/publications/
weighted_window_control.html>.
[gentle_RED] Floyd, S., "Recommendation on using [gentle_RED] Floyd, S., "Recommendation on using
the "gentle_" variant of RED", Web the "gentle_" variant of RED", Web
page , March 2000, <http:// page , March 2000, <http://
www.icir.org/floyd/red/gentle.html>. www.icir.org/floyd/red/gentle.html>.
[pBox] Floyd, S. and K. Fall, "Promoting the [pBox] Floyd, S. and K. Fall, "Promoting the
Use of End-to-End Congestion Control Use of End-to-End Congestion Control
in the Internet", IEEE/ACM in the Internet", IEEE/ACM
Transactions on Networking 7(4) 458-- Transactions on Networking 7(4) 458--
472, August 1999, <http:// 472, August 1999, <http://
www.aciri.org/floyd/ www.aciri.org/floyd/
end2end-paper.html>. end2end-paper.html>.
[pktByteEmail] Floyd, S., "RED: Discussions of Byte [pktByteEmail] Floyd, S., "RED: Discussions of Byte
and Packet Modes", Web page Red Queue and Packet Modes", email ,
Management, March 1997, <Available March 1997, <http://
at: http://ee.lbl.gov/floyd/ www-nrg.ee.lbl.gov/floyd/
REDaveraging.txt>. REDaveraging.txt>.
Appendix A. Idealised Wire Protocol Appendix A. Survey of RED Implementation Status
We will start by inventing an idealised congestion notification This Appendix is informative, not normative.
protocol before discussing how to make it practical. The idealised
protocol is shown to be correct using examples later in this
appendix.
A.1. Protocol Coding In May 2007 a survey was conducted of 84 vendors to assess how widely
drop probability based on packet size has been implemented in RED
Table 3. About 19% of those surveyed replied, giving a sample size
of 16. Although in most cases we do not have permission to identify
the respondents, we can say that those that have responded include
most of the larger equipment vendors, covering a large fraction of
the market. The two who gave permission to be identified were Cisco
and Alcatel-Lucent. The others range across the large network
equipment vendors at L3 & L2, firewall vendors, wireless equipment
vendors, as well as large software businesses with a small selection
of networking products. All those who responded confirmed that they
have not implemented the variant of RED with drop dependent on packet
size (2 were fairly sure they had not but needed to check more
thoroughly). At the time the survey was conducted, Linux did not
implement RED with packet-size bias of drop, although we have not
investigated a wider range of open source code.
Congestion notification involves the congested resource coding a +-------------------------------+----------------+-----------------+
congestion notification signal into the packet stream and the | Response | No. of vendors | %age of vendors |
transports decoding it. The idealised protocol uses two different +-------------------------------+----------------+-----------------+
(imaginary) fields in each datagram to signal congestion: one for | Not implemented | 14 | 17% |
byte congestion and one for packet congestion. | Not implemented (probably) | 2 | 2% |
| Implemented | 0 | 0% |
| No response | 68 | 81% |
| Total companies/orgs surveyed | 84 | 100% |
+-------------------------------+----------------+-----------------+
We are not saying two ECN fields will be needed (and we are not Table 3: Vendor Survey on byte-mode drop variant of RED (lower drop
saying that somehow a resource should be able to drop a packet in one probability for small packets)
of two different ways so that the transport can distinguish which
sort of drop it was!). These two congestion notification channels
are just a conceptual device. They allow us to defer having to
decide whether to distinguish between byte and packet congestion when
the network resource codes the signal or when the transport decodes
it.
However, although this idealised mechanism isn't intended for Where reasons have been given, the extra complexity of packet bias
implementation, we do want to emphasise that we may need to find a code has been most prevalent, though one vendor had a more principled
way to implement it, because it could become necessary to somehow reason for avoiding it--similar to the argument of this document.
distinguish between bit and packet congestion [RFC3714]. Currently,
packet-congestion is not the common case, but there is no guarantee
that it will not become common with future technology trends.
The idealised wire protocol is given below. It accounts for packet Our survey was of vendor implementations, so we cannot be certain
sizes at the transport layer, not in the network, and then only in about operator deployment. But we believe many queues in the
the case of bit-congestible resources. This avoids the perverse Internet are still tail-drop. The company of one of the co-authors
incentive to send smaller packets and the DoS vulnerability that (BT) has widely deployed RED, but many tail-drop queues are bound to
would otherwise result if the network were to bias towards them (see still exist, particularly in access network equipment and on
the motivating argument about avoiding perverse incentives in middleboxes like firewalls, where RED is not always available.
Section 3.3):
1. A packet-congestible resource trying to code congestion level p_p Routers using a memory architecture based on fixed size buffers with
into a packet stream should mark the idealised `packet borrowing may also still be prevalent in the Internet. As explained
congestion' field in each packet with probability p_p in Section 4.2.1, these also provide a marginal (but legitimate) bias
irrespective of the packet's size. The transport should then towards small packets. So even though RED byte-mode drop is not
take a packet with the packet congestion field marked to mean prevalent, it is likely there is still some bias towards small
just one mark, irrespective of the packet size. packets in the Internet due to tail drop and fixed buffer borrowing.
2. A bit-congestible resource trying to code time-varying byte- Appendix B. Sufficiency of Packet-Mode Drop
congestion level p_b into a packet stream should mark the `byte
congestion' field in each packet with probability p_b, again
irrespective of the packet's size. Unlike before, the transport
should take a packet with the byte congestion field marked to
count as a mark on each byte in the packet.
The worked examples in Appendix A.2 show that transports can extract This Appendix is informative, not normative.
sufficient and correct congestion notification from these protocols
for cases when two flows with different packet sizes have matching
bit rates or matching packet rates. Examples are also given that mix
these two flows into one to show that a flow with mixed packet sizes
would still be able to extract sufficient and correct information.
Sufficient and correct congestion information means that there is Here we check that packet-mode drop (or marking) in the network gives
sufficient information for the two different types of transport sufficiently generic information for the transport layer to use. We
requirements: check against a 2x2 matrix of four scenarios that may occur now or in
the future (Table 4). The horizontal and vertical dimensions have
been chosen because each tests extremes of sensitivity to packet size
in the transport and in the network respectively.
Ratio-based: Established transport congestion controls like TCP's Note that this section does not consider byte-mode drop at all.
[RFC5681] aim to achieve equal segment rates per RTT through the Having deprecated byte-mode drop, the goal here is to check that
same bottleneck--TCP friendliness [RFC3448]. They work with the packet-mode drop will be sufficient in all cases.
ratio of dropped to delivered segments (or marked to unmarked
segments in the case of ECN). The example scenarios show that
these ratio-based transports are effectively the same whether
counting in bytes or packets, because the units cancel out.
(Incidentally, this is why TCP's bit rate is still proportional to
packet size even when byte-counting is used, as recommended for
TCP in [RFC5681], mainly for orthogonal security reasons.)
Absolute-target-based: Other congestion controls proposed in the +-------------------------------+-----------------+-----------------+
research community aim to limit the volume of congestion caused to | Transport | a) Independent | b) Dependent on |
a constant weight parameter. [MulTCP][WindowPropFair] are | | of packet size | packet size of |
examples of weighted proportionally fair transports designed for | Network | of congestion | congestion |
cost-fair environments [Rate_fair_Dis]. In this case, the | | notifications | notifications |
transport requires a count (not a ratio) of dropped/marked bytes +-------------------------------+-----------------+-----------------+
in the bit-congestible case and of dropped/marked packets in the | 1) Predominantly | Scenario a1) | Scenario b1) |
packet congestible case. | bit-congestible network | | |
| 2) Mix of bit-congestible and | Scenario a2) | Scenario b2) |
| pkt-congestible network | | |
+-------------------------------+-----------------+-----------------+
A.2. Example Scenarios Table 4: Four Possible Congestion Scenarios
A.2.1. Notation Appendix B.1 focuses on the horizontal dimension of Table 4 checking
that packet-mode drop (or marking) gives sufficient information,
whether or not the transport uses it--scenarios b) and a)
respectively.
To prove our idealised wire protocol (Appendix A.1) is correct, we Appendix B.2 focuses on the vertical dimension of Table 4, checking
will compare two flows with different packet sizes, s_1 and s_2 [bit/ that packet-mode drop gives sufficient information to the transport
pkt], to make sure their transports each see the correct congestion whether resources in the network are bit-congestible or packet-
notification. Initially, within each flow we will take all packets congestible (these terms are defined in Section 1.1).
as having equal sizes, but later we will generalise to flows within
which packet sizes vary. A flow's bit rate, x [bit/s], is related to
its packet rate, u [pkt/s], by
x(t) = s.u(t).
We will consider a 2x2 matrix of four scenarios: Notation: To be concrete, we will compare two flows with different
packet sizes, s_1 and s_2. As an example, we will take s_1 = 60B
= 480b and s_2 = 1500B = 12,000b.
+-----------------------------+------------------+------------------+ A flow's bit rate, x [bps], is related to its packet rate, u
| resource type and | A) Equal bit | B) Equal pkt | [pps], by
| congestion level | rates | rates |
+-----------------------------+------------------+------------------+
| i) bit-congestible, p_b | (Ai) | (Bi) |
| ii) pkt-congestible, p_p | (Aii) | (Bii) |
+-----------------------------+------------------+------------------+
Table 3 x(t) = s.u(t).
A.2.2. Bit-congestible resource, equal bit rates (Ai) In the bit-congestible case, path congestion will be denoted by
p_b, and in the packet-congestible case by p_p. When either case
is implied, the letter p alone will denote path congestion.
Starting with the bit-congestible scenario, for two flows to maintain B.1. Packet-Size (In)Dependence in Transports
equal bit rates (Ai) the ratio of the packet rates must be the
inverse of the ratio of packet sizes: u_2/u_1 = s_1/s_2. So, for
instance, a flow of 60B packets would have to send 25x more packets
to achieve the same bit rate as a flow of 1500B packets. If a
congested resource marks proportion p_b of packets irrespective of
size, the ratio of marked packets received by each transport will
still be the same as the ratio of their packet rates, p_b.u_2/p_b.u_1
= s_1/s_2. So of the 25x more 60B packets sent, 25x more will be
marked than in the 1500B packet flow, but 25x more won't be marked
too.
In this scenario, the resource is bit-congestible, so it always uses In all cases we consider a packet-mode drop queue that indicates
our idealised bit-congestion field when it marks packets. Therefore congestion by dropping (or marking) packets with probability p
the transport should count marked bytes not packets. But it doesn't irrespective of packet size. We use an example value of loss
actually matter for ratio-based transports like TCP (Appendix A.1). (marking) probability, p=0.1%.
The ratio of marked to unmarked bytes seen by each flow will be p_b,
as will the ratio of marked to unmarked packets. Because they are
ratios, the units cancel out.
If a flow sent an inconsistent mixture of packet sizes, we have said A transport like RFC5681 TCP treats a congestion notification on any
it should count the ratio of marked and unmarked bytes not packets in packet whatever its size as one event. However, a network with just
order to correctly decode the level of congestion. But actually, if the packet-mode drop algorithm does give more information if the
all it is trying to do is decode p_b, it still doesn't matter. For transport chooses to use it. We will use Table 5 to illustrate this.
instance, imagine the two equal bit rate flows were actually one flow
at twice the bit rate sending a mixture of one 1500B packet for every
thirty 60B packets. 25x more small packets will be marked and 25x
more will be unmarked. The transport can still calculate p_b whether
it uses bytes or packets for the ratio. In general, for any
algorithm which works on a ratio of marks to non-marks, either bytes
or packets can be counted interchangeably, because the choice cancels
out in the ratio calculation.
However, where an absolute target rather than relative volume of We will set aside the last column until later. The columns labelled
congestion caused is important (Appendix A.1), as it is for "Flow 1" and "Flow 2" compare two flows consisting of 60B and 1500B
congestion accountability [Rate_fair_Dis], the transport must count packets respectively. The body of the table considers two separate
marked bytes not packets, in this bit-congestible case. Aside from cases, one where the flows have equal bit-rate and the other with
the goal of congestion accountability, this is how the bit rate of a equal packet-rates. In both cases, the two flows fill a 96Mbps link.
transport can be made independent of packet size; by ensuring the Therefore, in the equal bit-rate case they each have half the bit-
rate of congestion caused is kept to a constant weight rate (48Mbps). Whereas, with equal packet-rates, flow 1 uses 25
[WindowPropFair], rather than merely responding to the ratio of times smaller packets so it gets 25 times less bit-rate--it only gets
marked and unmarked bytes. 1/(1+25) of the link capacity (96Mbps/26 = 4Mbps after rounding). In
contrast flow 2 gets 25 times more bit-rate (92Mbps) in the equal
packet rate case because its packets are 25 times larger. The packet
rate shown for each flow could easily be derived once the bit-rate
was known by dividing bit-rate by packet size, as shown in the column
labelled "Formula".
Note the unit of byte-congestion-volume is the byte. Parameter Formula Flow 1 Flow 2 Combined
----------------------- ----------- ------- ------- --------
Packet size s/8 60B 1,500B (Mix)
Packet size s 480b 12,000b (Mix)
Pkt loss probability p 0.1% 0.1% 0.1%
A.2.3. Bit-congestible resource, equal packet rates (Bi) EQUAL BIT-RATE CASE
Bit-rate x 48Mbps 48Mbps 96Mbps
Packet-rate u = x/s 100kpps 4kpps 104kpps
Absolute pkt-loss-rate p*u 100pps 4pps 104pps
Absolute bit-loss-rate p*u*s 48kbps 48kbps 96kbps
Ratio of lost/sent pkts p*u/u 0.1% 0.1% 0.1%
Ratio of lost/sent bits p*u*s/(u*s) 0.1% 0.1% 0.1%
If two flows send different packet sizes but at the same packet rate, EQUAL PACKET-RATE CASE
their bit rates will be in the same ratio as their packet sizes, x_2/ Bit-rate x 4Mbps 92Mbps 96Mbps
x_1 = s_2/s_1. For instance, a flow sending 1500B packets at the Packet-rate u = x/s 8kpps 8kpps 15kpps
same packet rate as another sending 60B packets will be sending at Absolute pkt-loss-rate p*u 8pps 8pps 15pps
25x greater bit rate. In this case, if a congested resource marks Absolute bit-loss-rate p*u*s 4kbps 92kbps 96kbps
proportion p_b of packets irrespective of size, the ratio of packets Ratio of lost/sent pkts p*u/u 0.1% 0.1% 0.1%
received with the byte-congestion field marked by each transport will Ratio of lost/sent bits p*u*s/(u*s) 0.1% 0.1% 0.1%
be the same, p_b.u_2/p_b.u_1 = 1.
Because the byte-congestion field is marked, the transport should Table 5: Absolute Loss Rates and Loss Ratios for Flows of Small and
count marked bytes not packets. But because each flow sends Large Packets and Both Combined
consistently sized packets it still doesn't matter for ratio-based
transports. The ratio of marked to unmarked bytes seen by each flow
will be p_b, as will the ratio of marked to unmarked packets.
Therefore, if the congestion control algorithm is only concerned with
the ratio of marked to unmarked packets (as is TCP), both flows will
be able to decode p_b correctly whether they count packets or bytes.
But if the absolute volume of congestion is important, e.g. for So far we have merely set up the scenarios. We now consider
congestion accountability, the transport must count marked bytes not congestion notification in the scenario. Two TCP flows with the same
packets. Then the lower bit rate flow using smaller packets will round trip time aim to equalise their packet-loss-rates over time.
rightly be perceived as causing less byte-congestion even though its That is the number of packets lost in a second, which is the packets
packet rate is the same. per second (u) multiplied by the probability that each one is dropped
(p). Thus TCP converges on the "Equal packet-rate" case, where both
flows aim for the same "Absolute packet-loss-rate" (both 8pps in the
table).
If the two flows are mixed into one, of bit rate x1+x2, with equal Packet-mode drop actually gives flows sufficient information to
packet rates of each size packet, the ratio p_b will still be measure their loss-rate in bits per second, if they choose, not just
measurable by counting the ratio of marked to unmarked bytes (or packets per second. Each flow can count the size of a lost or marked
packets because the ratio cancels out the units). However, if the packet and scale its rate-response in proportion (as TFRC-SP does).
absolute volume of congestion is required, the transport must count The result is shown in the row entitled "Absolute bit-loss-rate",
the sum of congestion marked bytes, which indeed gives a correct where the bits lost in a second is the packets per second (u)
measure of the rate of byte-congestion p_b(x_1 + x_2) caused by the multiplied by the probability of losing a packet (p) multiplied by
combined bit rate. the packet size (s). Such an algorithm would try to remove any
imbalance in bit-loss-rate such as the wide disparity in the "Equal
packet-rate" case (4kbps vs. 92kbps). Instead, a packet-size-
dependent algorithm would aim for equal bit-loss-rates, which would
drive both flows towards the "Equal bit-rate" case, by driving them
to equal bit-loss-rates (both 48kbps in this example).
A.2.4. Pkt-congestible resource, equal bit rates (Aii) The explanation so far has assumed that each flow consists of packets
of only one constant size. Nonetheless, it extends naturally to
flows with mixed packet sizes. In the right-most column of Table 5 a
flow of mixed size packets is created simply by considering flow 1
and flow 2 as a single aggregated flow. There is no need for a flow
to maintain an average packet size. It is only necessary for the
transport to scale its response to each congestion indication by the
size of each individual lost (or marked) packet. Taking for example
the "Equal packet-rate" case, in one second about 8 small packets and
8 large packets are lost (making closer to 15 than 16 losses per
second due to rounding). If the transport multiplies each loss by
its size, in one second it responds to 8*480b and 8*12,000b lost
bits, adding up to 96,000 lost bits in a second. This double checks
correctly, being the same as 0.1% of the total bit-rate of 96Mbps.
For completeness, the formula for absolute bit-loss-rate is p(u1*s1+
u2*s2).
Moving to the case of packet-congestible resources, we now take two Incidentally, a transport will always measure the loss probability
flows that send different packet sizes at the same bit rate, but this the same irrespective of whether it measures in packets or in bytes.
time the pkt-congestion field is marked by the resource with In other words, the ratio of lost to sent packets will be the same as
probability p_p. As in scenario Ai with the same bit rates but a the ratio of lost to sent bytes. (This is why TCP's bit rate is
bit-congestible resource, the flow with smaller packets will have a still proportional to packet size even when byte-counting is used, as
higher packet rate, so more packets will be both marked and unmarked, recommended for TCP in [RFC5681], mainly for orthogonal security
but in the same proportion. reasons.) This is intuitively obvious by comparing two example
flows; one with 60B packets, the other with 1500B packets. If both
flows pass through a queue with drop probability 0.1%, each flow will
lose 1 in 1,000 packets. In the stream of 60B packets the ratio of
bytes lost to sent will be 60B in every 60,000B; and in the stream of
1500B packets, the loss ratio will be 1,500B out of 1,500,000B. When
the transport responds to the ratio of lost to sent packets, it will
measure the same ratio whether it measures in packets or bytes: 0.1%
in both cases. The fact that this ratio is the same whether measured
in packets or bytes can be seen in Table 5, where the ratio of lost
to sent packets and the ratio of lost to sent bytes is always 0.1% in
all cases (recall that the scenario was set up with p=0.1%).
This time, the transport should only count marks without taking into This discussion of how the ratio can be measured in packets or bytes
account packet sizes. Transports will get the same result, p_p, by is only raised here to highlight that it is irrelevant to this memo!
decoding the ratio of marked to unmarked packets in either flow. Whether a transport depends on packet size or not depends on how this
ratio is used within the congestion control algorithm.
If one flow imitates the two flows but merged together, the bit rate So far we have shown that packet-mode drop passes sufficient
will double with more small packets than large. The ratio of marked information to the transport layer so that the transport can take
to unmarked packets will still be p_p. But if the absolute number of account of bit-congestion, by using the sizes of the packets that
pkt-congestion marked packets is counted it will accumulate at the indicate congestion. We have also shown that the transport can
combined packet rate times the marking probability, p_p(u_1+u_2), 26x choose not to take packet size into account if it wishes. We will
faster than packet congestion accumulates in the single 1500B packet now consider whether the transport can know which to do.
flow of our example, as required.
But if the transport is interested in the absolute number of packet B.2. Bit-Congestible and Packet-Congestible Indications
congestion, it should just count how many marked packets arrive. For
instance, a flow sending 60B packets will see 25x more marked packets
than one sending 1500B packets at the same bit rate, because it is
sending more packets through a packet-congestible resource.
Note the unit of packet congestion is a packet. As a thought-experiment, imagine an idealised congestion notification
protocol that supports both bit-congestible and packet-congestible
resources. It would require at least two ECN flags, one for each of
bit-congestible and packet-congestible resources.
A.2.5. Pkt-congestible resource, equal packet rates (Bii) 1. A packet-congestible resource trying to code congestion level p_p
into a packet stream should mark the idealised `packet
congestion' field in each packet with probability p_p
irrespective of the packet's size. The transport should then
take a packet with the packet congestion field marked to mean
just one mark, irrespective of the packet size.
Finally, if two flows with the same packet rate, pass through a 2. A bit-congestible resource trying to code time-varying byte-
packet-congestible resource, they will both suffer the same congestion level p_b into a packet stream should mark the `byte
proportion of marking, p_p, irrespective of their packet sizes. On congestion' field in each packet with probability p_b, again
detecting that the pkt-congestion field is marked, the transport irrespective of the packet's size. Unlike before, the transport
should count packets, and it will be able to extract the ratio p_p of should take a packet with the byte congestion field marked to
marked to unmarked packets from both flows, irrespective of packet count as a mark on each byte in the packet.
sizes.
Even if the transport is monitoring the absolute amount of packets This hides a fundamental problem--much more fundamental than whether
congestion over a period, still it will see the same amount of packet we can magically create header space for yet another ECN flag, or
congestion from either flow. whether it would work while being deployed incrementally.
Distinguishing drop from delivery naturally provides just one
implicit bit of congestion indication information--the packet is
either dropped or not. It is hard to drop a packet in two ways that
are distinguishable remotely. This is a similar problem to that of
distinguishing wireless transmission losses from congestive losses.
And if the two equal packet rates of different size packets are mixed This problem would not be solved even if ECN were universally
together in one flow, the packet rate will double, so the absolute deployed. A congestion notification protocol must survive a
volume of packet-congestion will accumulate at twice the rate of transition from low levels of congestion to high. Marking two states
either flow, 2p_p.u_1 = p_p(u_1+u_2). is feasible with explicit marking, but much harder if packets are
dropped. Also, it will not always be cost-effective to implement AQM
at every low level resource, so drop will often have to suffice.
Appendix B. Byte-mode Drop Complicates Policing Congestion Response We are not saying two ECN fields will be needed (and we are not
saying that somehow a resource should be able to drop a packet in one
of two different ways so that the transport can distinguish which
sort of drop it was!). These two congestion notification channels
are a conceptual device to illustrate a dilemma we could face in the
future. Section 3 gives four good reasons why it would be a bad idea
to allow for packet size by biasing drop probability in favour of
small packets within the network. The impracticality of our thought
experiment shows that it will be hard to give transports a practical
way to know whether to take account of the size of congestion
indication packets or not.
This appendix explains why the ability of networks to police the Fortunately, this dilemma is not pressing because by design most
response of _any_ transport to congestion depends on bit-congestible equipment becomes bit-congested before its packet-processing becomes
network resources only doing packet-mode not byte-mode drop. congested (as already outlined in Section 1.1). Therefore transports
can be designed on the relatively sound assumption that a congestion
indication will usually imply bit-congestion.
To be able to police a transport's response to congestion when Nonetheless, although the above idealised protocol isn't intended for
fairness can only be judged over time and over all an individual's implementation, we do want to emphasise that research is needed to
flows, the policer has to have an integrated view of all the predict whether there are good reasons to believe that packet
congestion an individual (not just one flow) has caused due to all congestion might become more common, and if so, to find a way to
traffic entering the Internet from that individual. This is termed somehow distinguish between bit and packet congestion [RFC3714].
congestion accountability.
But a byte-mode drop algorithm has to depend on the local MTU of the Recently, the dual resource queue (DRQ) proposal [DRQ] has been made
line - an algorithm needs to use some concept of a 'normal' packet on the premise that, as network processors become more cost
size. Therefore, one dropped or marked packet is not necessarily effective, per packet operations will become more complex
equivalent to another unless you know the MTU at the queue where it (irrespective of whether more function in the network is desirable).
was dropped/marked. To have an integrated view of a user, we believe Consequently the premise is that CPU congestion will become more
congestion policing has to be located at an individual's attachment common. DRQ is a proposed modification to the RED algorithm that
point to the Internet [I-D.ietf-conex-concepts-uses]. But from there folds both bit congestion and packet congestion into one signal
it cannot know the MTU of each remote queue that caused each drop/ (either loss or ECN).
mark. Therefore it cannot take an integrated approach to policing
all the responses to congestion of all the transports of one
individual. Therefore it cannot police anything.
The security/incentive argument _for_ packet-mode drop is similar. Finally, we note one further complication. Strictly, packet-
Firstly, confining RED to packet-mode drop would not preclude congestible resources are often cycle-congestible. For instance, for
bottleneck policing approaches such as [pBox] as it seems likely they routing look-ups load depends on the complexity of each look-up and
could work just as well by monitoring the volume of dropped bytes whether the pattern of arrivals is amenable to caching or not. This
rather than packets. Secondly packet-mode dropping/marking naturally also reminds us that any solution must not require a forwarding
allows the congestion notification of packets to be globally engine to use excessive processor cycles in order to decide how to
meaningful without relying on MTU information held elsewhere. say it has no spare processor cycles.
Because we recommend that a dropped/marked packet should be taken to Appendix C. Byte-mode Drop Complicates Policing Congestion Response
mean that all the bytes in the packet are dropped/marked, a policer
can remain robust against bits being re-divided into different size
packets or across different size flows [Rate_fair_Dis]. Therefore
policing would work naturally with just simple packet-mode drop in
RED.
In summary, making drop probability depend on the size of the packets There are two main classes of approach to policing congestion
that bits happen to be divided into simply encourages the bits to be response: i) policing at each bottleneck link or ii) policing at the
divided into smaller packets. Byte-mode drop would therefore edges of networks. Packet-mode drop in RED is compatible with
irreversibly complicate any attempt to fix the Internet's incentive either, while byte-mode drop precludes edge policing.
structures.
Appendix C. Changes from Previous Versions The simplicity of an edge policer relies on one dropped or marked
packet being equivalent to another of the same size without having to
know which link the drop or mark occurred at. However, the byte-mode
drop algorithm has to depend on the local MTU of the line--it needs
to use some concept of a 'normal' packet size. Therefore, one
dropped or marked packet from a byte-mode drop algorithm is not
necessarily equivalent to another from a different link. A policing
function local to the link can know the local MTU where the
congestion occurred. However, a policer at the edge of the network
cannot, at least not without a lot of complexity.
The early research proposals for type (i) policing at a bottleneck
link [pBox] used byte-mode drop, then detected flows that contributed
disproportionately to the number of packets dropped. However, with
no extra complexity, later proposals used packet mode drop and looked
for flows that contributed a disproportionate amount of dropped bytes
[CHOKe_Var_Pkt].
Work is progressing on the congestion exposure protocol (ConEx
[I-D.ietf-conex-concepts-uses]), which enables a type (ii) edge
policer located at a user's attachment point. The idea is to be able
to take an integrated view of the effect of all a user's traffic on
any link in the internetwork. However, byte-mode drop would
effectively preclude such edge policing because of the MTU issue
above.
Indeed, making drop probability depend on the size of the packets
that bits happen to be divided into would simply encourage the bits
to be divided into smaller packets in order to confuse policing. In
contrast, as long as a dropped/marked packet is taken to mean that
all the bytes in the packet are dropped/marked, a policer can remain
robust against bits being re-divided into different size packets or
across different size flows [Rate_fair_Dis].
Appendix D. Changes from Previous Versions
To be removed by the RFC Editor on publication. To be removed by the RFC Editor on publication.
Full incremental diffs between each version are available at Full incremental diffs between each version are available at
<http://www.cs.ucl.ac.uk/staff/B.Briscoe/pubs.html#byte-pkt-congest>
or
<http://tools.ietf.org/wg/tsvwg/draft-ietf-tsvwg-byte-pkt-congest/> <http://tools.ietf.org/wg/tsvwg/draft-ietf-tsvwg-byte-pkt-congest/>
(courtesy of the rfcdiff tool): (courtesy of the rfcdiff tool):
From -04 to -05:
* Changed from Informational to BCP and highlighted non-normative
sections and appendices
* Removed language about consensus
* Added "Example Comparing Packet-Mode Drop and Byte-Mode Drop"
* Arranged "Motivating Arguments" into a more logical order and
completely rewrote "Transport-Independent Network" & "Scaling
Congestion Control with Packet Size" arguments. Removed "Why
Now?"
* Clarified applicability of certain recommendations
* Shifted vendor survey to an Appendix
* Cut down "Outstanding Issues and Next Steps"
* Re-drafted the start of the conclusions to highlight the three
distinct areas of concern
* Completely re-wrote appendices
* Editorial corrections throughout.
From -03 to -04: From -03 to -04:
* Reordered Sections 2 and 3, and some clarifications here and * Reordered Sections 2 and 3, and some clarifications here and
there based on feedback from Colin Perkins and Mirja there based on feedback from Colin Perkins and Mirja
Kuehlewind. Kuehlewind.
From -02 to -03 (this version) From -02 to -03 (this version)
* Structural changes: * Structural changes:
skipping to change at page 41, line 11 skipping to change at page 41, line 31
* Clarified the last point about why this is a good time to sort * Clarified the last point about why this is a good time to sort
out this issue: because it will be hard / impossible to design out this issue: because it will be hard / impossible to design
new transports unless we decide whether the network or the new transports unless we decide whether the network or the
transport is allowing for packet size. transport is allowing for packet size.
* Added statement explaining the horizon of the memo is long * Added statement explaining the horizon of the memo is long
term, but with short term expediency in mind. term, but with short term expediency in mind.
* Added material on scaling congestion control with packet size * Added material on scaling congestion control with packet size
(Section 3.1). (Section 3.4).
* Separated out issue of normalising TCP's bit rate from issue of * Separated out issue of normalising TCP's bit rate from issue of
preference to control packets (Section 3.4). preference to control packets (Section 3.2).
* Divided up Congestion Measurement section for clarity, * Divided up Congestion Measurement section for clarity,
including new material on fixed size packet buffers and buffer including new material on fixed size packet buffers and buffer
carving (Section 4.1.1 & Section 4.2.1) and on congestion carving (Section 4.1.1 & Section 4.2.1) and on congestion
measurement in wireless link technologies without queues measurement in wireless link technologies without queues
(Section 4.1.2). (Section 4.1.2).
* Added section on 'Making Transports Robust against Control * Added section on 'Making Transports Robust against Control
Packet Losses' (Section 4.2.3) with existing & new material Packet Losses' (Section 4.2.3) with existing & new material
included. included.
* Added tabulated results of vendor survey on byte-mode drop * Added tabulated results of vendor survey on byte-mode drop
variant of RED (Table 2). variant of RED (Table 3).
From -00 to -01: From -00 to -01:
* Clarified applicability to drop as well as ECN. * Clarified applicability to drop as well as ECN.
* Highlighted DoS vulnerability. * Highlighted DoS vulnerability.
* Emphasised that drop-tail suffers from similar problems to * Emphasised that drop-tail suffers from similar problems to
byte-mode drop, so only byte-mode drop should be turned off, byte-mode drop, so only byte-mode drop should be turned off,
not RED itself. not RED itself.
 End of changes. 159 change blocks. 
823 lines changed or deleted 864 lines changed or added

This html diff was produced by rfcdiff 1.41. The latest version is available from http://tools.ietf.org/tools/rfcdiff/