draft-ietf-quic-manageability-00.txt   draft-ietf-quic-manageability-01.txt 
Network Working Group M. Kuehlewind Network Working Group M. Kuehlewind
Internet-Draft B. Trammell Internet-Draft B. Trammell
Intended status: Informational ETH Zurich Intended status: Informational ETH Zurich
Expires: January 4, 2018 D. Druta Expires: April 28, 2018 October 25, 2017
AT&T
July 03, 2017
Manageability of the QUIC Transport Protocol Manageability of the QUIC Transport Protocol
draft-ietf-quic-manageability-00 draft-ietf-quic-manageability-01
Abstract Abstract
This document discusses manageability of the QUIC transport protocol, This document discusses manageability of the QUIC transport protocol,
focusing on caveats impacting network operations involving QUIC focusing on caveats impacting network operations involving QUIC
traffic. Its intended audience is network operators, as well as traffic. Its intended audience is network operators, as well as
content providers that rely on the use of QUIC-aware middleboxes, content providers that rely on the use of QUIC-aware middleboxes,
e.g. for load balancing. e.g. for load balancing.
Status of This Memo Status of This Memo
skipping to change at page 1, line 36 skipping to change at page 1, line 34
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 January 4, 2018. This Internet-Draft will expire on April 28, 2018.
Copyright Notice Copyright Notice
Copyright (c) 2017 IETF Trust and the persons identified as the Copyright (c) 2017 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
skipping to change at page 2, line 12 skipping to change at page 2, line 10
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 . . . . . . . . . . . . . . . . . . . . . . . . 2 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1. Notational Conventions . . . . . . . . . . . . . . . . . 3 1.1. Notational Conventions . . . . . . . . . . . . . . . . . 3
2. Features of the QUIC Wire Image . . . . . . . . . . . . . . . 3 2. Features of the QUIC Wire Image . . . . . . . . . . . . . . . 3
2.1. QUIC Packet Header Structure . . . . . . . . . . . . . . 3 2.1. QUIC Packet Header Structure . . . . . . . . . . . . . . 4
2.2. Integrity Protection of the Wire Image . . . . . . . . . 5 2.2. Integrity Protection of the Wire Image . . . . . . . . . 5
2.3. Connection ID and Rebinding . . . . . . . . . . . . . . . 5 2.3. Connection ID and Rebinding . . . . . . . . . . . . . . . 5
2.4. Packet Numbers . . . . . . . . . . . . . . . . . . . . . 5 2.4. Packet Numbers . . . . . . . . . . . . . . . . . . . . . 6
2.5. Initial Handshake and PMTUD . . . . . . . . . . . . . . . 6 2.5. Initial Handshake and PMTUD . . . . . . . . . . . . . . . 6
2.6. Version Negotiation and Greasing . . . . . . . . . . . . 6 2.6. Version Negotiation and Greasing . . . . . . . . . . . . 6
3. Specific Network Management Tasks . . . . . . . . . . . . . . 6 3. Network-visible information about QUIC flows . . . . . . . . 6
3.1. Stateful Treatment of QUIC Traffic . . . . . . . . . . . 6 3.1. Identifying QUIC traffic . . . . . . . . . . . . . . . . 7
3.2. Measurement of QUIC Traffic . . . . . . . . . . . . . . . 7 3.1.1. Identifying Negotiated Version . . . . . . . . . . . 7
3.3. DDoS Detection and Mitigation . . . . . . . . . . . . . . 8 3.1.2. Rejection of Garbage Traffic . . . . . . . . . . . . 7
3.4. QoS support and ECMP . . . . . . . . . . . . . . . . . . 8 3.2. Connection confirmation . . . . . . . . . . . . . . . . . 7
3.5. Load Balancing using the Connection ID . . . . . . . . . 9 3.3. Flow association . . . . . . . . . . . . . . . . . . . . 8
4. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 10 3.4. Flow teardown . . . . . . . . . . . . . . . . . . . . . . 8
5. Security Considerations . . . . . . . . . . . . . . . . . . . 10 3.5. Round-trip time measurement . . . . . . . . . . . . . . . 8
6. Contributors . . . . . . . . . . . . . . . . . . . . . . . . 10 3.6. Packet loss measurement . . . . . . . . . . . . . . . . . 9
7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 10 3.7. Flow symmetry measurement . . . . . . . . . . . . . . . . 9
8. References . . . . . . . . . . . . . . . . . . . . . . . . . 11 4. Specific Network Management Tasks . . . . . . . . . . . . . . 9
8.1. Normative References . . . . . . . . . . . . . . . . . . 11 4.1. Stateful treatment of QUIC traffic . . . . . . . . . . . 9
8.2. Informative References . . . . . . . . . . . . . . . . . 11 4.2. Passive network performance measurement and
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 12 troubleshooting . . . . . . . . . . . . . . . . . . . . . 9
4.3. Server cooperation with load balancers . . . . . . . . . 10
4.4. DDoS Detection and Mitigation . . . . . . . . . . . . . . 10
4.5. QoS support and ECMP . . . . . . . . . . . . . . . . . . 11
5. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 11
6. Security Considerations . . . . . . . . . . . . . . . . . . . 11
7. Contributors . . . . . . . . . . . . . . . . . . . . . . . . 12
8. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 12
9. References . . . . . . . . . . . . . . . . . . . . . . . . . 12
9.1. Normative References . . . . . . . . . . . . . . . . . . 12
9.2. Informative References . . . . . . . . . . . . . . . . . 12
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 13
1. Introduction 1. Introduction
QUIC [QUIC] is a new transport protocol currently under development QUIC [QUIC] is a new transport protocol currently under development
in the IETF quic working group, focusing on support of semantics as in the IETF quic working group, focusing on support of semantics as
needed for HTTP/2 [QUIC-HTTP]. Based on current deployment needed for HTTP/2 [QUIC-HTTP]. Based on current deployment
practices, QUIC is encapsulated in UDP and encrypted by default. The practices, QUIC is encapsulated in UDP and encrypted by default. The
current version of QUIC integrates TLS [QUIC-TLS] to encrypt all current version of QUIC integrates TLS [QUIC-TLS] to encrypt all
payload data and most control information. Given QUIC is an end-to- payload data and most control information. Given QUIC is an end-to-
end transport protocol, all information in the protocol header, even end transport protocol, all information in the protocol header, even
skipping to change at page 3, line 50 skipping to change at page 4, line 12
This document is focused on the protocol as presently defined in This document is focused on the protocol as presently defined in
[QUIC] and [QUIC-TLS], and will change to track those documents. [QUIC] and [QUIC-TLS], and will change to track those documents.
2.1. QUIC Packet Header Structure 2.1. QUIC Packet Header Structure
The QUIC packet header is under active development; see section 5 of The QUIC packet header is under active development; see section 5 of
[QUIC] for the present header structure. [QUIC] for the present header structure.
The first bit of the QUIC header indicates the present of a long The first bit of the QUIC header indicates the present of a long
header that exposes more information than the short header. The long header that exposes more information than the short header. The long
header is typically used during connection start or for other control header is used during connection start including version negotiation,
processes while the short header will be used on most data packets to server retry, and 0-RTT data while the short header is used after the
limited unnecessary header overhead. The fields and location of handshake and therefore on most data packets to limited unnecessary
these fields as defined by the current version of QUIC for the long header overhead. The fields and location of these fields as defined
header are fixed for all future version as well. However, note that by the current version of QUIC for the long header are fixed for all
future versions of QUIC may provide additional fields. In the future version as well. However, note that future versions of QUIC
current version of quic the long header for all header types has a may provide additional fields. In the current version of quic the
fixed length, containing, besides the Header Form bit, a 7-bit header long header for all header types has a fixed length, containing,
Type, a 64-bit Connection ID, a 32-bit Packet Number, and a 32-bit besides the Header Form bit, a 7-bit header Type, a 64-bit Connection
Version. The short header is variable length where bits after the ID, a 32-bit Packet Number, and a 32-bit Version. The short header
Header Form bit indicate the present on the Connection ID, and the is variable length where bits after the Header Form bit indicate the
length of the packet number. present on the Connection ID, and the length of the packet number.
The following information may be exposed in the packet header: The following information may be exposed in the packet header:
o header type: the long header has a 7-bit header type field o header type: the long header has a 7-bit header type field
following the Header Form bit. The current version of QUIC following the Header Form bit. The current version of QUIC
defines 7 header types, namely Version Negotiation, Client defines 6 header types, namely Version Negotiation, Client
Initial, Server Stateless Retry, Server Cleartext, Client Initial, Server Stateless Retry, Server Cleartext, Client
Cleartext, 0-RTT Protected, 1-RTT Protected (key phase 0), 1-RTT Cleartext, 0-RTT Protected.
Protected (key phase 1), and Public Reset.
o connection ID: The connection ID is always present on the long and o connection ID: The connection ID is always present on the long and
optionally present on the short header indicated by the Connection optionally present on the short header indicated by the Connection
ID Flag. If present at the short header it at the same position ID Flag. If present at the short header it at the same position
then for the long header. The position and length pf the then for the long header. The position and length pf the
congestion ID itself as well as the Connection ID flag in the congestion ID itself as well as the Connection ID flag in the
short header is fixed for all versions of QUIC. The connection ID short header is fixed for all versions of QUIC. The connection ID
identifies the connection associated with a QUIC packet, for load- identifies the connection associated with a QUIC packet, for load-
balancing and NAT rebinding purposes; see Section 3.5 and balancing and NAT rebinding purposes; see Section 4.3 and
Section 2.3. Therefore it is also expected that the Connection ID Section 2.3. Therefore it is also expected that the Connection ID
will either be present on all packets of a flow or none of the will either be present on all packets of a flow or none of the
short header packets. However, this field is under endpoint short header packets. However, this field is under endpoint
control and there is protocol mechanism that hinders the sending control and there is no protocol mechanism that hinders the
endpoint to revise its decision about exposing the Connection ID sending endpoint to revise its decision about exposing the
at any time during the connection. Connection ID at any time during the connection.
o packet number: Every packet has an associated packet number. The o packet number: Every packet has an associated packet number. The
packet number increases with each packet, and the least- packet number increases with each packet, and the least-
significant bits of the packet number are present on each packet. significant bits of the packet number are present on each packet.
In the short header the length of the exposed packet number field In the short header the length of the exposed packet number field
is defined by the (short) header type and can either be 8, 16, or is defined by the (short) header type and can either be 8, 16, or
32 bits. See Section 2.4. 32 bits. See Section 2.4.
o version number: The version number is present on the long headers o version number: The version number is present on the long headers
and identifies the version used for that packet, expect for the and identifies the version used for that packet, expect for the
Version negotiation packet. The version negotiation packet is Version negotiation packet. The version negotiation packet is
fixed for all version of QUIC and contains a lit of versions that fixed for all version of QUIC and contains a list of versions that
is supported by the sender. However the version in the version is supported by the sender. The version in the version field of
field of the header is the reflected version of the clients the Version Negotiation packet is the reflected version of the
initial packet and is therefore explicitly not supported by the Client Initial packet and is therefore explicitly n ot supported
sender. by the sender.
o key phase: The short header further has a Key Phase flag that is o key phase: The short header further has a Key Phase flag that is
used by the endpoint identify the right key that was used to used by the endpoint identify the right key that was used to
encrypt the packet. Different key phases are indicated with the encrypt the packet.
use of the long header by using to different header types for
protected long header packets.
2.2. Integrity Protection of the Wire Image 2.2. Integrity Protection of the Wire Image
As soon as the cryptographic context is established, all information As soon as the cryptographic context is established, all information
in the QUIC header, including those exposed in the packet header, is in the QUIC header, including those exposed in the packet header, is
integrity protected. Further, information that were sent and exposed integrity protected. Further, information that were sent and exposed
in previous packets when the cryptographic context was established in previous packets when the cryptographic context was established
yet, e.g. for the cryptographic initial handshake itself, will be yet, e.g. for the cryptographic initial handshake itself, will be
validate later during the cryptographic handshake, such as the validated later during the cryptographic handshake, such as the
version number. Therefore, devices on path MUST NOT change any version number. Therefore, devices on path MUST NOT change any
information or bits in QUIC packet headers. As alteration of header information or bits in QUIC packet headers. As alteration of header
information would cause packet drop due to a failed integrity check information would cause packet drop due to a failed integrity check
at the receiver, or can even lead to connection termination. at the receiver, or can even lead to connection termination.
2.3. Connection ID and Rebinding 2.3. Connection ID and Rebinding
The connection ID in the QUIC packer header is used to allow routing The connection ID in the QUIC packer header is used to allow routing
of QUIC packets at load balancers on other than five-tuple of QUIC packets at load balancers on other than five-tuple
information, ensuring that related flows are appropriately balanced information, ensuring that related flows are appropriately balanced
together; and to allow rebinding of a connection after one of the together; and to allow rebinding of a connection after one of the
endpoint's addresses changes - usually the client's, in the case of endpoint's addresses changes - usually the client's, in the case of
the HTTP binding. The connection ID is proposed by the server during the HTTP binding. The client set a Connection ID in the Initial
connection establishment, and a server might provide additional Client packet that will be used during the handshake. A new
connection IDs that can the used by the client at any time during the connection ID is then provided by the server during connection
connection. Therefore if a flow changes one of its IP addresses it establishment, that will be used in the short header after the
may keep the same connection ID, or the connection ID may also change handshake. Further a server might provide additional connection IDs
together with the IP address migration, avoiding linkability; see that can the used by the client at any time during the connection.
Section 7.6 of [QUIC]. Therefore if a flow changes one of its IP addresses it may keep the
same connection ID, or the connection ID may also change together
with the IP address migration, avoiding linkability; see Section 7.6
of [QUIC].
2.4. Packet Numbers 2.4. Packet Numbers
The packet number field is always present in the QUIC packet header. The packet number field is always present in the QUIC packet header.
The packet number exposes the least significant 32, 16, or 8 bits of The packet number exposes the least significant 32, 16, or 8 bits of
an internal packet counter per flow direction that increments with an internal packet counter per flow direction that increments with
each packet sent. This packet counter is initialized with a random each packet sent. This packet counter is initialized with a random
31-bit initial value at the start of a connection. 31-bit initial value at the start of a connection.
Unlike TCP sequence numbers, this packet number increases with every Unlike TCP sequence numbers, this packet number increases with every
skipping to change at page 6, line 36 skipping to change at page 6, line 48
Also note that the list of versions in the Version Negotiation packet Also note that the list of versions in the Version Negotiation packet
may contain reserved versions. This mechanism is used to avoid may contain reserved versions. This mechanism is used to avoid
ossification in the implementation on the selection mechanism. ossification in the implementation on the selection mechanism.
Further, a client may send a Initial Client packet with a reserved Further, a client may send a Initial Client packet with a reserved
version number to trigger version negotiation. In the Version version number to trigger version negotiation. In the Version
Negotiation packet the connection ID and packet number of the Client Negotiation packet the connection ID and packet number of the Client
Initial packet are reflected to provide a proof of return- Initial packet are reflected to provide a proof of return-
routability. Therefore changing these information will also cause routability. Therefore changing these information will also cause
the connection to fail. the connection to fail.
3. Specific Network Management Tasks 3. Network-visible information about QUIC flows
In this section, we address specific network management and This section addresses the different kinds of observations and
measurement techniques and how QUIC's design impacts them. inferences that can be made about QUIC flows by a passive observer in
the network based on the wire image in Section 2. Here we assume a
bidirectional observer (one that can see packets in both directions
in the sequence in which they are carried on the wire) unless noted.
3.1. Stateful Treatment of QUIC Traffic 3.1. Identifying QUIC traffic
Stateful network devices such as firewalls use exposed header The QUIC wire image is not specifically designed to be
information to support state setup and tear-down. [STATEFULNESS] distinguishable from other UDP traffic.
provides a general model for in-network state management on these
devices, independent of transport protocol. Features already present
in QUIC may be used for state maintenance in this model. Here, there
are two important goals: distinguishing valid QUIC connection
establishment from other traffic, in order to establish state; and
determining the end of a QUIC connection, in order to tear that state
down.
Both, 1-RTT and O-RTT connection establishment, using a TLS handshake The only application binding currently defined for QUIC is HTTP
on stream 0, is detectable using heuristics similar to those used to [QUIC-HTTP]. HTTP over QUIC uses UDP port 443 by default, although
detect TLS over TCP. 0-RTT connection may additional also send data URLs referring to resources available over HTTP over QUIC may specify
packets, right after the client hello. These data may be reorder in alternate port numbers. Simple assumptions about whether a given
the network, therefore it may be possible that 0-RTT Protected data flow is using QUIC based upon a UDP port number may therefore not
packet are seen before the Client Initial packet. hold; see also [RFC7605] section 5.
Exposure of connection shutdown is currently under discussion; see 3.1.1. Identifying Negotiated Version
https://github.com/quicwg/base-drafts/issues/353 and
https://github.com/quicwg/base-drafts/pull/20.
3.2. Measurement of QUIC Traffic An in-network observer assuming that a set of packets belongs to a
QUIC flow can infer the version number in use by observing the
handshake: a Client Initial with a given version followed by Server
Cleartext packet with the same version implies acceptance of that
version.
Passive measurement of TCP performance parameters is commonly Negotiated version cannot be identified for flows for which a
deployed in access and enterprise networks to aid troubleshooting and handshake is not observed, such as in the case of NAT rebinding;
performance monitoring without requiring the generation of active however, these flows can be associated with flows for which a version
measurement traffic. has been identified; see Section 3.3.
The presence of packet numbers on all QUIC packets allows the trivial In the rest of this section, we discuss only packets belonging to
one-sided estimation of packet loss and reordering between the sender Version 1 QUIC flows, and assume that these packets have been
and a given observation point. However, since retransmissions are identified as such through the observation of a version negotiation.
not identifiable as such, loss between an observation point and the
receiver cannot be reliably estimated. 3.1.2. Rejection of Garbage Traffic
A related question is whether a first packet of a given flow on known
QUIC-associated port is a valid QUIC packet, in order to support in-
network filtering of garbage UDP packets (reflection attacks, random
backscatter). While heuristics based on the first byte of the packet
(packet type) could be used to separate valid from invalid first
packet types, the deployment of such heuristics is not recommended,
as packet types may have different meanings in future versions of the
protocol.
3.2. Connection confirmation
Connection establishment requires cleartext packets and is using a
TLS handshake on stream 0. Therefore it is detectable using
heuristics similar to those used to detect TLS over TCP. 0-RTT
connection may additional also send data packets, right after the
Client Initial with the TLS client hello. These data may be
reordered in the network, therefore it may be possible that 0-RTT
Protected data packets are seen before the Client Initial packet.
3.3. Flow association
The QUIC Connection ID (see Section 2.3) is designed to allow an on-
path device such as a load-balancer to associate two flows as
identified by five-tuple when the address and port of one of the
endpoints changes; e.g. due to NAT rebinding or server IP address
migration. An observer keeping flow state can associate a connection
ID with a given flow, and can associate a known flow with a new flow
when when observing a packet sharing a connection ID and one endpoint
address (IP address and port) with the known flow.
The connection ID to be used for a long-running flow is chosen by the
server (see [QUIC] section 5.6) during the handshake. This value
should be treated as opaque; see Section 4.3 for caveats regarding
connection ID selection at servers.
3.4. Flow teardown
The QUIC does not expose the end of a connection; the only indication
to on-path devices that a flow has ended is that packets are no
longer observed. Stateful devices on path such as NATs and firewalls
must therefore use idle timeouts to determine when to drop state for
QUIC flows.
Changes to this behavior are currently under discussion: see
https://github.com/quicwg/base-drafts/issues/602.
3.5. Round-trip time measurement
Round-trip time of QUIC flows can be inferred by observation once per
flow, during the handshake, as in passive TCP measurement. The delay
between the Client Initial packet and the Server Cleartext packet
sent back to the client represents the RTT component on the path
between the observer and the server, and the delay between this
packet and the Client Cleartext packet in reply represents the RTT
component on the path between the observer and the client. This
measurement necessarily includes any application delay at both sides.
Note that the Server's reply mayalso be a Version Negotiation or
Server Stateless Retry packet. In this case the Client will send
another Client Initial or the connection will fail.
The lack of any acknowledgement information or timestamping The lack of any acknowledgement information or timestamping
information in the QUIC packet header makes running passive information in the QUIC wire image makes running passive RTT
estimation of latency via round trip time (RTT) impossible. RTT can estimation impossible.
only be measured at connection establishment time, by observing the
Client Initial packet and the Server's reply to this packet which
maybe a Server Cleartext, Version Negotiation, or Server Stateless
Retry packet.
Note that adding a packet number echo (as in Changes to this behavior are currently under discussion: see
https://github.com/quicwg/base-drafts/pull/367 or https://github.com/quicwg/base-drafts/issues/631.
https://github.com/quicwg/base-drafts/pull/368) to the public header
would allow passive RTT measurement at on-path observation points.
For efficiency purposes, this packet number echo need not be carried
on every packet, and could be made optional, allowing endpoints to
make a measurability/efficiency tradeoff; see section 4 of [IPIM].
Note further that this facility would have significantly better
measurability characteristics than sequence-acknowledgement-based RTT
measurement currently available in TCP on typical asymmetric flows,
as adequate samples will be available in both directions, and packet
number echo would be decoupled from the underlying acknowledgment
machinery; see e.g. [Ding2015]
Note in-network devices can inspect and correlate connection IDs for 3.6. Packet loss measurement
partial tracking of mobility events.
3.3. DDoS Detection and Mitigation All QUIC packets carry packet numbers in cleartext, and while the
protocol allows packet numbers to be skipped, skipping is not
recommended in the general case. This allows the trivial one-sided
estimation of packet loss and reordering between the sender and a
given observation point ("upstream loss"). However, since
retransmissions are not identifiable as such, loss between an
observation point and the receiver ("downstream loss") cannot be
reliably estimated.
For enterprises and network operators one of the biggest management 3.7. Flow symmetry measurement
challenges is dealing with Distributed Denial of Service (DDoS)
attacks. Some network operators offer Security as a Service (SaaS)
solutions that detect attacks by monitoring, analyzing and filtering
traffic. These approaches generally utilize network flow data
[RFC7011]. If any flows pose a threat, usually they are routed to a
"scrubbing environment" where the traffic is filtered, allowing the
remaining "good" traffic to continue to the customer environment.
This type of DDoS mitigation is fundamentally based on tracking state QUIC explicitly exposes which side of a connection is a client and
for flows (see Section 3.1) that have receiver confirmation and a which side is a server during the handshake. In addition, the
proof of return-routability, and classifying flows as legitimate or symmerty of a flow (whether primarily client-to-server, primarily
DoS traffic. The QUIC packet header currently does not support an server-to-client, or roughly bidirectional, as input to basic traffic
explicit mechanism to easily distinguish legitimate QUIC traffic from classification techniques) can be inferred through the measurement of
other UDP traffic. However, the first packet in a QUIC connection data rate in each direction. While QUIC traffic is protected and
will usually be a Client Initial packet. This can be used to ACKS may be padded, padding is not required.
identify the first packet of the connection.
If the QUIC handshake was not observed by the defense system, the 4. Specific Network Management Tasks
connection ID can be used as a confirmation signal as per
[STATEFULNESS] if present in both directions.
Further, the use of a connection ID to support connection migration In this section, we address specific network management and
measurement techniques and how QUIC's design impacts them.
4.1. Stateful treatment of QUIC traffic
Stateful treatment of QUIC traffic is possible through QUIC traffic
and version identification (Section 3.1) and observation of the
handshake for connection confirmation (Section 3.2). The lack of any
visible end-of-flow signal (Section 3.4) means that this state must
be purged either through timers or through least-recently-used
eviction, depending on application requirements.
4.2. Passive network performance measurement and troubleshooting
Extremely limited loss and RTT measurement are possible by passive
observation of QUIC traffic; see Section 3.5 and Section 3.6.
4.3. Server cooperation with load balancers
In the case of content distribution networking architectures
including load balancers, the connection ID provides a way for the
server to signal information about the desired treatment of a flow to
the load balancers.
Server-generated Connection IDs must not encode any information other
that that needed to route packets to the appropriate backend
server(s): typically the identity of the backend server or pool of
servers, if the data-center's load balancing system keeps "local"
state of all flows itself. Care must be exercised to ensure that the
information encoded in the Connection ID is not sufficient to
identify unique end users. Note that by encoding routing information
in the Connection ID, load balancers open up a new attack vector that
allows bad actors to direct traffic at a specific backend server or
pool. It is therefore recommended that Server-Generated Connection
ID includes a cryptographic MAC that the load balancer pool server
are able to identify and discard packets featuring an invalid MAC.
4.4. DDoS Detection and Mitigation
Current practices in detection and mitigation of Distributed Denial
of Service (DDoS) attacks generally involve passive measurement using
network flow data [RFC7011], classification of traffic into "good"
(productive) and "bad" (DoS) flows, and filtering of these bad flows
in a "scrubbing" environment. Key to successful DDoS mitigation is
efficient classification of this traffic.
Limited first-packet garbage detection as in Section 3.1.2 and
stateful tracking of QUIC traffic as in Section 4.1 above can be used
in this classification step. For traffic where the classification
step did not observe a QUIC handshake, the presence of packets
carrying the same Connection ID in both directions is a further
indication of legitimate traffic. Note that these classification
techniques help only against floods of garbage traffic, not against
DDoS attacks using legitimate QUIC clients.
Note that the use of a connection ID to support connection migration
renders 5-tuple based filtering insufficient, and requires more state renders 5-tuple based filtering insufficient, and requires more state
to be maintained by DDoS defense systems. However, it is to be maintained by DDoS defense systems. However, it is
questionable if connection migrations needs to be supported in a DDOS questionable if connection migrations needs to be supported in a DDOS
attack. If the connection migration is not visible to the network attack. If the connection migration is not visible to the network
that performs the DDoS detection, an active, migrated QUIC connection that performs the DDoS detection, an active, migrated QUIC connection
may be blocked by such a system under attack. However, a defense may be blocked by such a system under attack. However, a defense
system might simply rely on the fast resumption mechanism provided by system might simply rely on the fast resumption mechanism provided by
QUIC. This problem is also related to these issues under discussion: QUIC. See also https://github.com/quicwg/base-drafts/issues/203
https://github.com/quicwg/base-drafts/issues/203
3.4. QoS support and ECMP 4.5. QoS support and ECMP
[EDITOR'S NOTE: this is a bit speculative; keep?]
QUIC does not provide any additional information on requirements on QUIC does not provide any additional information on requirements on
Quality of Service (QoS) provided from the network. QUIC assumes Quality of Service (QoS) provided from the network. QUIC assumes
that all packets with the same 5-tuple {dest addr, source addr, that all packets with the same 5-tuple {dest addr, source addr,
protocol, dest port, source port} will receive similar network protocol, dest port, source port} will receive similar network
treatment. That means all stream that are multiplexed over the same treatment. That means all stream that are multiplexed over the same
QUIC connection require the same network treatment and are handled by QUIC connection require the same network treatment and are handled by
the same congestion controller. If differential network treatment is the same congestion controller. If differential network treatment is
desired, multiple QUIC connection to the same server might be used, desired, multiple QUIC connections to the same server might be used,
given that establishing a new connection using 0-RTT support is cheap given that establishing a new connection using 0-RTT support is cheap
and fast. and fast.
QoS mechanisms in the network MAY also use the connection ID for QoS mechanisms in the network MAY also use the connection ID for
service differentiation as usually a change of connection ID is bind service differentiation, as a change of connection ID is bound to a
to a change of address which anyway is likely to lead to a re-route change of address which anyway is likely to lead to a re-route on a
on a different path with different network characteristics. different path with different network characteristics.
Given that QUIC is more tolerant of packet re-ordering than TCP (see Given that QUIC is more tolerant of packet re-ordering than TCP (see
Section 2.4), Equal-cost multi-path routing (ECMP) does not Section 2.4), Equal-cost multi-path routing (ECMP) does not
necessarily need to be flow based. However, 5-tuple (plus eventually necessarily need to be flow based. However, 5-tuple (plus eventually
connection ID if present) matching is still beneficial for QoS given connection ID if present) matching is still beneficial for QoS given
all packets are handled by the same congestion controller. all packets are handled by the same congestion controller.
3.5. Load Balancing using the Connection ID 5. IANA Considerations
The Connection ID is used in part to support load balancing in
content distribution networks (CDNs), which operate complex,
geographically distributed pools of back-end servers, fronted by load
balancing systems. These load balancers are responsible for
identifying the most appropriate server for each connection and for
routing all packets belonging to that connection to the chosen
server.
Load balancers are often deployed in pools for redundancy and load
sharing. For high availability, it is important that when packets
belonging to a flow start to arrive at a different load balancer in
the load balancer pool, the packets continue to be forwarded to the
original server in the server pool.
Support for seamless connection migration is an important design goal
of QUIC - a necessity due to the proliferation of mobile connected
devices. This connection persistence provides an additional
challenge for multi-homed anycast-based services often employed by
large content owners and CDNs. The challenge is that a migration to
a different network in the middle of the connection greatly increases
the chances of the packets routed to a different anycast point of
presence (POP) due to the new network's different connectivity and
Internet peering arrangements. The load balancer in the new POP,
potentially thousands of miles away, will not have information about
the established flow and would not be able to route it back to the
original POP.
Load balancers may cooperate with servers or server pools behind them
to use a server-generated Connection ID value, in order to support
stateless load balancing, even across NAT rebinding or other address
change events (see Section 2.3). See section 5.7 of [QUIC].
Server-generated Connection IDs must not encode any information other
that that needed to route packets to the appropriate backend
server(s): typically the identity of the backend server or pool of
servers, if the data-center's load balancing system keeps "local"
state of all flows itself. Care must be exercised to ensure that the
information encoded in the Connection ID is not sufficient to
identify unique end users. Note that by encoding routing information
in the Connection ID, load balancers open up a new attack vector that
allows bad actors to direct traffic at a specific backend server or
pool. It is therefore recommended that Server-Generated Connection
ID includes a cryptographic MAC that the load balancer pool server
are able to identify and discard packets featuring an invalid MAC.
4. IANA Considerations
This document has no actions for IANA. This document has no actions for IANA.
5. Security Considerations 6. Security Considerations
Supporting manageability of QUIC traffic inherently involves Supporting manageability of QUIC traffic inherently involves
tradeoffs with the confidentiality of QUIC's control information; tradeoffs with the confidentiality of QUIC's control information;
this entire document is therefore security-relevant. this entire document is therefore security-relevant.
Some of the properties of the QUIC header used in network management Some of the properties of the QUIC header used in network management
are irrelevant to application-layer protocol operation and/or user are irrelevant to application-layer protocol operation and/or user
privacy. For example, packet number exposure (and echo, as proposed privacy. For example, packet number exposure (and echo, as proposed
in this document), as well as connection establishment exposure for in this document), as well as connection establishment exposure for
1-RTT establishment, make no additional information about user 1-RTT establishment, make no additional information about user
traffic available to devices on path. traffic available to devices on path.
At the other extreme, supporting current traffic classification At the other extreme, supporting current traffic classification
methods that operate through the deep packet inspection (DPI) of methods that operate through the deep packet inspection (DPI) of
application-layer headers are directly antithetical to QUIC's goal to application-layer headers are directly antithetical to QUIC's goal to
provide confidentiality to its application-layer protocol(s); in provide confidentiality to its application-layer protocol(s); in
these cases, alternatives must be found. these cases, alternatives must be found.
6. Contributors 7. Contributors
Igor Lubashev contributed text to Section 3.5 on the use of the Dan Druta contributed text to Section 4.4. Igor Lubashev contributed
connection ID for load balancing. text to Section 4.3 on the use of the connection ID for load
balancing.
7. Acknowledgments 8. Acknowledgments
This work is partially supported by the European Commission under This work is partially supported by the European Commission under
Horizon 2020 grant agreement no. 688421 Measurement and Architecture Horizon 2020 grant agreement no. 688421 Measurement and Architecture
for a Middleboxed Internet (MAMI), and by the Swiss State Secretariat for a Middleboxed Internet (MAMI), and by the Swiss State Secretariat
for Education, Research, and Innovation under contract no. 15.0268. for Education, Research, and Innovation under contract no. 15.0268.
This support does not imply endorsement. This support does not imply endorsement.
8. References 9. References
8.1. Normative References
[QUIC] Iyengar, J. and M. Thomson, "QUIC: A UDP-Based Multiplexed
and Secure Transport", draft-ietf-quic-transport-04 (work
in progress), June 2017.
[QUIC-HTTP]
Bishop, M., "Hypertext Transfer Protocol (HTTP) over
QUIC", draft-ietf-quic-http-04 (work in progress), June
2017.
[QUIC-TLS] 9.1. Normative References
Thomson, M. and S. Turner, "Using Transport Layer Security
(TLS) to Secure QUIC", draft-ietf-quic-tls-04 (work in
progress), June 2017.
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/
DOI 10.17487/RFC2119, March 1997, RFC2119, March 1997, <https://www.rfc-editor.org/info/
<http://www.rfc-editor.org/info/rfc2119>. rfc2119>.
8.2. Informative References 9.2. Informative References
[Ding2015] [Ding2015]
Ding, H. and M. Rabinovich, "TCP Stretch Acknowledgments Ding, H. and M. Rabinovich, "TCP Stretch Acknowledgments
and Timestamps - Findings and Impliciations for Passive and Timestamps - Findings and Impliciations for Passive
RTT Measurement (ACM Computer Communication Review)", July RTT Measurement (ACM Computer Communication Review)", July
2015, <http://www.sigcomm.org/sites/default/files/ccr/ 2015, <http://www.sigcomm.org/sites/default/files/ccr/
papers/2015/July/0000000-0000002.pdf>. papers/2015/July/0000000-0000002.pdf>.
[IPIM] Allman, M., Beverly, R., and B. Trammell, "In-Protocol [IPIM] Allman, M., Beverly, R., and B. Trammell, "In-Protocol
Internet Measurement (arXiv preprint 1612.02902)", Internet Measurement (arXiv preprint 1612.02902)",
December 2016, <https://arxiv.org/abs/1612.02902>. December 2016, <https://arxiv.org/abs/1612.02902>.
[QUIC] Iyengar, J. and M. Thomson, "QUIC: A UDP-Based Multiplexed
and Secure Transport", draft-ietf-quic-transport-07 (work
in progress), October 2017.
[QUIC-HTTP]
Bishop, M., "Hypertext Transfer Protocol (HTTP) over
QUIC", draft-ietf-quic-http-07 (work in progress), October
2017.
[QUIC-TLS]
Thomson, M. and S. Turner, "Using Transport Layer Security
(TLS) to Secure QUIC", draft-ietf-quic-tls-07 (work in
progress), October 2017.
[RFC7011] Claise, B., Ed., Trammell, B., Ed., and P. Aitken, [RFC7011] Claise, B., Ed., Trammell, B., Ed., and P. Aitken,
"Specification of the IP Flow Information Export (IPFIX) "Specification of the IP Flow Information Export (IPFIX)
Protocol for the Exchange of Flow Information", STD 77, Protocol for the Exchange of Flow Information", STD 77,
RFC 7011, DOI 10.17487/RFC7011, September 2013, RFC 7011, DOI 10.17487/RFC7011, September 2013,
<http://www.rfc-editor.org/info/rfc7011>. <https://www.rfc-editor.org/info/rfc7011>.
[STATEFULNESS] [RFC7605] Touch, J., "Recommendations on Using Assigned Transport
Kuehlewind, M., Trammell, B., and J. Hildebrand, Port Numbers", BCP 165, RFC 7605, DOI 10.17487/RFC7605,
"Transport-Independent Path Layer State Management", August 2015, <https://www.rfc-editor.org/info/rfc7605>.
draft-trammell-plus-statefulness-03 (work in progress),
March 2017.
Authors' Addresses Authors' Addresses
Mirja Kuehlewind Mirja Kuehlewind
ETH Zurich ETH Zurich
Gloriastrasse 35 Gloriastrasse 35
8092 Zurich 8092 Zurich
Switzerland Switzerland
Email: mirja.kuehlewind@tik.ee.ethz.ch Email: mirja.kuehlewind@tik.ee.ethz.ch
Brian Trammell Brian Trammell
ETH Zurich ETH Zurich
Gloriastrasse 35 Gloriastrasse 35
8092 Zurich 8092 Zurich
Switzerland Switzerland
Email: ietf@trammell.ch Email: ietf@trammell.ch
Dan Druta
AT&T
Email: dd5826@att.com
 End of changes. 49 change blocks. 
216 lines changed or deleted 272 lines changed or added

This html diff was produced by rfcdiff 1.46. The latest version is available from http://tools.ietf.org/tools/rfcdiff/