draft-ietf-mptcp-architecture-05.txt   rfc6182.txt 
Internet Engineering Task Force A. Ford Internet Engineering Task Force (IETF) A. Ford
Internet-Draft Roke Manor Research Request for Comments: 6182 Roke Manor Research
Intended status: Informational C. Raiciu Category: Informational C. Raiciu
Expires: July 25, 2011 M. Handley ISSN: 2070-1721 M. Handley
University College London University College London
S. Barre S. Barre
Universite catholique de Universite catholique de Louvain
Louvain
J. Iyengar J. Iyengar
Franklin and Marshall College Franklin and Marshall College
January 21, 2011 March 2011
Architectural Guidelines for Multipath TCP Development Architectural Guidelines for Multipath TCP Development
draft-ietf-mptcp-architecture-05
Abstract Abstract
Hosts are often connected by multiple paths, but TCP restricts Hosts are often connected by multiple paths, but TCP restricts
communications to a single path per transport connection. Resource communications to a single path per transport connection. Resource
usage within the network would be more efficient were these multiple usage within the network would be more efficient were these multiple
paths able to be used concurrently. This should enhance user paths able to be used concurrently. This should enhance user
experience through improved resilience to network failure and higher experience through improved resilience to network failure and higher
throughput. throughput.
This document outlines architectural guidelines for the development This document outlines architectural guidelines for the development
of a Multipath Transport Protocol, with references to how these of a Multipath Transport Protocol, with references to how these
architectural components come together in the development of a architectural components come together in the development of a
Multipath TCP protocol. This document lists certain high level Multipath TCP (MPTCP). This document lists certain high-level design
design decisions that provide foundations for the design of the MPTCP decisions that provide foundations for the design of the MPTCP
protocol, based upon these architectural requirements. protocol, based upon these architectural requirements.
Status of this Memo Status of This Memo
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering This document is not an Internet Standards Track specification; it is
Task Force (IETF). Note that other groups may also distribute published for informational purposes.
working documents as Internet-Drafts. The list of current Internet-
Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months This document is a product of the Internet Engineering Task Force
and may be updated, replaced, or obsoleted by other documents at any (IETF). It represents the consensus of the IETF community. It has
time. It is inappropriate to use Internet-Drafts as reference received public review and has been approved for publication by the
material or to cite them other than as "work in progress." Internet Engineering Steering Group (IESG). Not all documents
approved by the IESG are a candidate for any level of Internet
Standard; see Section 2 of RFC 5741.
This Internet-Draft will expire on July 25, 2011. Information about the current status of this document, any errata,
and how to provide feedback on it may be obtained at
http://www.rfc-editor.org/info/rfc6182.
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 . . . . . . . . . . . . . . . . . . . . . . . . . 4 1. Introduction ....................................................4
1.1. Requirements Language . . . . . . . . . . . . . . . . . . 5 1.1. Requirements Language ......................................5
1.2. Terminology . . . . . . . . . . . . . . . . . . . . . . . 5 1.2. Terminology ................................................5
1.3. Reference Scenario . . . . . . . . . . . . . . . . . . . . 6 1.3. Reference Scenario .........................................6
2. Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2. Goals ...........................................................6
2.1. Functional Goals . . . . . . . . . . . . . . . . . . . . . 6 2.1. Functional Goals ...........................................6
2.2. Compatibility Goals . . . . . . . . . . . . . . . . . . . 7 2.2. Compatibility Goals ........................................7
2.2.1. Application Compatibility . . . . . . . . . . . . . . 7 2.2.1. Application Compatibility ...........................7
2.2.2. Network Compatibility . . . . . . . . . . . . . . . . 8 2.2.2. Network Compatibility ...............................8
2.2.3. Compatibility with other network users . . . . . . . . 9 2.2.3. Compatibility with Other Network Users .............10
2.3. Security Goals . . . . . . . . . . . . . . . . . . . . . . 10 2.3. Security Goals ............................................10
2.4. Related Protocols . . . . . . . . . . . . . . . . . . . . 10 2.4. Related Protocols .........................................10
3. An Architectural Basis For Multipath TCP . . . . . . . . . . . 10 3. An Architectural Basis for Multipath TCP .......................11
4. A Functional Decomposition of MPTCP . . . . . . . . . . . . . 12 4. A Functional Decomposition of MPTCP ............................12
5. High-Level Design Decisions . . . . . . . . . . . . . . . . . 14 5. High-Level Design Decisions ....................................14
5.1. Sequence Numbering . . . . . . . . . . . . . . . . . . . . 14 5.1. Sequence Numbering ........................................14
5.2. Reliability and Retransmissions . . . . . . . . . . . . . 15 5.2. Reliability and Retransmissions ...........................15
5.3. Buffers . . . . . . . . . . . . . . . . . . . . . . . . . 17 5.3. Buffers ...................................................17
5.4. Signalling . . . . . . . . . . . . . . . . . . . . . . . . 18 5.4. Signaling .................................................18
5.5. Path Management . . . . . . . . . . . . . . . . . . . . . 19 5.5. Path Management ...........................................19
5.6. Connection Identification . . . . . . . . . . . . . . . . 20 5.6. Connection Identification .................................20
5.7. Congestion Control . . . . . . . . . . . . . . . . . . . . 21 5.7. Congestion Control ........................................21
5.8. Security . . . . . . . . . . . . . . . . . . . . . . . . . 21 5.8. Security ..................................................21
6. Software Interactions . . . . . . . . . . . . . . . . . . . . 22 6. Software Interactions ..........................................23
6.1. Interactions with Applications . . . . . . . . . . . . . . 22 6.1. Interactions with Applications ............................23
6.2. Interactions with Management Systems . . . . . . . . . . . 23 6.2. Interactions with Management Systems ......................23
7. Interactions with Middleboxes . . . . . . . . . . . . . . . . 23 7. Interactions with Middleboxes ..................................23
8. Contributors . . . . . . . . . . . . . . . . . . . . . . . . . 25 8. Contributors ...................................................25
9. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 25 9. Acknowledgements ...............................................25
10. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 25 10. Security Considerations .......................................26
11. Security Considerations . . . . . . . . . . . . . . . . . . . 25 11. References ....................................................26
12. References . . . . . . . . . . . . . . . . . . . . . . . . . . 26 11.1. Normative References .....................................26
12.1. Normative References . . . . . . . . . . . . . . . . . . . 26 11.2. Informative References ...................................26
12.2. Informative References . . . . . . . . . . . . . . . . . . 26
Appendix A. Changelog . . . . . . . . . . . . . . . . . . . . . . 28
A.1. Changes since draft-ietf-mptcp-architecture-04 . . . . . . 28
A.2. Changes since draft-ietf-mptcp-architecture-03 . . . . . . 28
A.3. Changes since draft-ietf-mptcp-architecture-02 . . . . . . 28
A.4. Changes since draft-ietf-mptcp-architecture-01 . . . . . . 28
A.5. Changes since draft-ietf-mptcp-architecture-00 . . . . . . 28
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 28
1. Introduction 1. Introduction
As the Internet evolves, demands on Internet resources are ever- As the Internet evolves, demands on Internet resources are ever-
increasing, but often these resources (in particular, bandwidth) increasing, but often these resources (in particular, bandwidth)
cannot be fully utilised due to protocol constraints both on the end- cannot be fully utilized due to protocol constraints both on the end-
systems and within the network. If these resources could be used systems and within the network. If these resources could be used
concurrently, end user experience could be greatly improved. Such concurrently, end user experience could be greatly improved. Such
enhancements would also reduce the necessary expenditure on network enhancements would also reduce the necessary expenditure on network
infrastructure that would otherwise be needed to create an equivalent infrastructure that would otherwise be needed to create an equivalent
improvement in user experience. By the application of resource improvement in user experience. By the application of resource
pooling [3], these available resources can be 'pooled' such that they pooling [3], these available resources can be 'pooled' such that they
appear as a single logical resource to the user. appear as a single logical resource to the user.
Multipath transport aims to realize some of the goals of resource Multipath transport aims to realize some of the goals of resource
pooling by simultaneously making use of multiple disjoint (or pooling by simultaneously making use of multiple disjoint (or
partially disjoint) paths across a network. The two key benefits of partially disjoint) paths across a network. The two key benefits of
multipath transport are: multipath transport are the following:
o To increase the resilience of the connectivity by providing o To increase the resilience of the connectivity by providing
multiple paths, protecting end hosts from the failure of one. multiple paths, protecting end hosts from the failure of one.
o To increase the efficiency of the resource usage, and thus o To increase the efficiency of the resource usage, and thus
increase the network capacity available to end hosts. increase the network capacity available to end hosts.
Multipath TCP is a modified version of TCP [1] that implements a Multipath TCP is a modified version of TCP [1] that implements a
multipath transport and achieves these goals by pooling multiple multipath transport and achieves these goals by pooling multiple
paths within a transport connection, transparently to the paths within a transport connection, transparently to the
application. Multipath TCP is primarily concerned with utilising application. Multipath TCP is primarily concerned with utilizing
multiple paths end-to-end, where one or both end host is multi-homed. multiple paths end-to-end, where one or both of the end hosts are
It may also have applications where multiple paths exist within the multihomed. It may also have applications where multiple paths exist
network and can be manipulated by an end host, such as using within the network and can be manipulated by an end host, such as
different port numbers with ECMP [4]. using different port numbers with Equal Cost MultiPath (ECMP) [4].
MPTCP, defined in [5], is a specific protocol that instantiates the MPTCP, defined in [5], is a specific protocol that instantiates the
Multipath TCP concept. This document looks both at general Multipath TCP concept. This document looks both at general
architectural principles for a Multipath TCP fulfilling the goals architectural principles for a Multipath TCP fulfilling the goals
described in Section 2, as well as the key design decisions behind described in Section 2, as well as the key design decisions behind
MPTCP, which are detailed in Section 5. MPTCP, which are detailed in Section 5.
Although multihoming and multipath functions are not new to transport Although multihoming and multipath functions are not new to transport
protocols (SCTP [6] being a notable example), MPTCP aims to gain protocols (Stream Control Transmission Protocol (SCTP) [6] being a
wide-scale deployment by recognising the importance of application notable example), MPTCP aims to gain wide-scale deployment by
and network compatibility goals. These goals, discussed in detail in recognizing the importance of application and network compatibility
Section 2, relate to the appearance of MPTCP to the network (so non- goals. These goals, discussed in detail in Section 2, relate to the
MPTCP-aware entities see it as TCP) and to the application (through appearance of MPTCP to the network (so non-MPTCP-aware entities see
providing an service equivalent to TCP for non-MPTCP-aware it as TCP) and to the application (through providing a service
applications). equivalent to TCP for non-MPTCP-aware applications).
This document has three key purposes: (i) it describes goals for a This document has three key purposes: (i) it describes goals for a
multipath transport - goals that MPTCP is designed to meet; (ii) it multipath transport -- goals that MPTCP is designed to meet; (ii) it
lays out an architectural basis for MPTCP's design - a discussion lays out an architectural basis for MPTCP's design -- a discussion
that applies to other multipath transports as well; and (iii) it that applies to other multipath transports as well; and (iii) it
discusses and documents high-level design decisions made in MPTCP's discusses and documents high-level design decisions made in MPTCP's
development, and considers their implications. development, and considers their implications.
Companion documents to this architectural overview are those which Companion documents to this architectural overview are those that
provide details of the protocol extensions [5], congestion control provide details of the protocol extensions [5], congestion control
algorithms [7], and application-level considerations [8]. Put algorithms [7], and application-level considerations [8]. Put
together, these components specify a complete Multipath TCP design. together, these components specify a complete Multipath TCP design.
We note that specific components are replaceable in accordance with Note that specific components are replaceable in accordance with the
the layer and functional decompositions discussed in this document. layer and functional decompositions discussed in this document.
1.1. Requirements Language 1.1. Requirements Language
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 RFC 2119 [2]. document are to be interpreted as described in RFC 2119 [2].
1.2. Terminology 1.2. Terminology
Regular/Single-Path TCP: The standard version of the TCP [1] Regular/Single-Path TCP: The standard version of TCP [1] in use
protocol in use today, operating between a single pair of IP today, operating between a single pair of IP addresses and ports.
addresses.
Multipath TCP: A modified version of the TCP protocol that supports Multipath TCP: A modified version of the TCP protocol that supports
the simultaneous use of multiple paths between hosts. the simultaneous use of multiple paths between hosts.
Path: A sequence of links between a sender and a receiver, defined Path: A sequence of links between a sender and a receiver, defined
in this context by a source and destination address pair. in this context by a source and destination address pair.
Host: An end host either initiating or terminating a Multipath TCP Host: An end host either initiating or terminating a Multipath TCP
connection. connection.
skipping to change at page 6, line 9 skipping to change at page 6, line 9
which forms part of a larger Multipath TCP connection. which forms part of a larger Multipath TCP connection.
(Multipath TCP) Connection: A set of one or more subflows combined (Multipath TCP) Connection: A set of one or more subflows combined
to provide a single Multipath TCP service to an application at a to provide a single Multipath TCP service to an application at a
host. host.
1.3. Reference Scenario 1.3. Reference Scenario
The diagram shown in Figure 1 illustrates a typical usage scenario The diagram shown in Figure 1 illustrates a typical usage scenario
for Multipath TCP. Two hosts, A and B, are communicating with each for Multipath TCP. Two hosts, A and B, are communicating with each
other. These hosts are multi-homed and multi-addressed, providing other. These hosts are multihomed and multi-addressed, providing two
two disjoint connections to the Internet. The addresses on each host disjoint connections to the Internet. The addresses on each host are
are referred to as A1, A2, B1 and B2. There are therefore up to four referred to as A1, A2, B1, and B2. There are therefore up to four
different paths between the two hosts: A1-B1, A1-B2, A2-B1, A2-B2. different paths between the two hosts: A1-B1, A1-B2, A2-B1, A2-B2.
+------+ __________ +------+ +------+ __________ +------+
| |A1 ______ ( ) ______ B1| | | |A1 ______ ( ) ______ B1| |
| Host |--/ ( ) \--| Host | | Host |--/ ( ) \--| Host |
| | ( Internet ) | | | | ( Internet ) | |
| A |--\______( )______/--| B | | A |--\______( )______/--| B |
| |A2 (__________) B2| | | |A2 (__________) B2| |
+------+ +------+ +------+ +------+
Figure 1: Simple Multipath TCP Usage Scenario Figure 1: Simple Multipath TCP Usage Scenario
The scenario could have any number of addresses (1 or more) on each The scenario could have any number of addresses (1 or more) on each
host, as long as the number of paths available between the two hosts host, as long as the number of paths available between the two hosts
is 2 or more (i.e. num_addr(A) * num_addr(B) > 1). The paths created is 2 or more (i.e., num_addr(A) * num_addr(B) > 1). The paths
by these address combinations through the Internet need not be created by these address combinations through the Internet need not
entirely disjoint - potential fairness issues introduced by shared be entirely disjoint -- potential fairness issues introduced by
bottlenecks need to be handled by the Multipath TCP congestion shared bottlenecks need to be handled by the Multipath TCP congestion
controller. Furthermore, the paths through the Internet often do not controller. Furthermore, the paths through the Internet often do not
provide a pure end-to-end service, and instead may be affected by provide a pure end-to-end service, and instead may be affected by
middleboxes such as NATs and Firewalls. middleboxes such as NATs and firewalls.
2. Goals 2. Goals
This section outlines primary goals that Multipath TCP aims to meet. This section outlines primary goals that Multipath TCP aims to meet.
These are broadly broken down into: functional goals, which steer These are broadly broken down into the following: functional goals,
services and features that Multipath TCP must provide; and which steer services and features that Multipath TCP must provide,
compatibility goals, which determine how Multipath TCP should appear and compatibility goals, which determine how Multipath TCP should
to entities that interact with it. appear to entities that interact with it.
2.1. Functional Goals 2.1. Functional Goals
In supporting the use of multiple paths, Multipath TCP has the In supporting the use of multiple paths, Multipath TCP has the
following two functional goals. following two functional goals.
o Improve Throughput: Multipath TCP MUST support the concurrent use o Improve Throughput: Multipath TCP MUST support the concurrent use
of multiple paths. To meet the minimum performance incentives for of multiple paths. To meet the minimum performance incentives for
deployment, a Multipath TCP connection over multiple paths SHOULD deployment, a Multipath TCP connection over multiple paths SHOULD
achieve no lesser throughput than a single TCP connection over the achieve no worse throughput than a single TCP connection over the
best constituent path. best constituent path.
o Improve Resilience: Multipath TCP MUST support the use of multiple o Improve Resilience: Multipath TCP MUST support the use of multiple
paths interchangeably for resilience purposes, by permitting paths interchangeably for resilience purposes, by permitting
segments to be sent and re-sent on any available path. It follows segments to be sent and re-sent on any available path. It follows
that, in the worst case, the protocol MUST be no less resilient that, in the worst case, the protocol MUST be no less resilient
than regular single-path TCP. than regular single-path TCP.
As distribution of traffic among available paths and responses to As distribution of traffic among available paths and responses to
congestion are done in accordance with resource pooling principles congestion are done in accordance with resource pooling principles
skipping to change at page 7, line 29 skipping to change at page 7, line 29
its use. A host supporting Multipath TCP that requires the other its use. A host supporting Multipath TCP that requires the other
host to do so too must be able to detect reliably whether this host host to do so too must be able to detect reliably whether this host
does in fact support the required extensions, using them if so, and does in fact support the required extensions, using them if so, and
otherwise automatically falling back to single-path TCP. otherwise automatically falling back to single-path TCP.
2.2. Compatibility Goals 2.2. Compatibility Goals
In addition to the functional goals listed above, a Multipath TCP In addition to the functional goals listed above, a Multipath TCP
must meet a number of compatibility goals in order to support must meet a number of compatibility goals in order to support
deployment in today's Internet. These goals fall into the following deployment in today's Internet. These goals fall into the following
categories: categories.
2.2.1. Application Compatibility 2.2.1. Application Compatibility
Application compatibility refers to the appearance of Multipath TCP Application compatibility refers to the appearance of Multipath TCP
to the application both in terms of the API that can be used and the to the application both in terms of the API that can be used and the
expected service model that is provided. expected service model that is provided.
Multipath TCP MUST follow the same service model as TCP [1]: in- Multipath TCP MUST follow the same service model as TCP [1]: in-
order, reliable, and byte-oriented delivery. Furthermore, a order, reliable, and byte-oriented delivery. Furthermore, a
Multipath TCP connection SHOULD provide the application with no worse Multipath TCP connection SHOULD provide the application with no worse
throughput or resilience than it would expect from running a single throughput or resilience than it would expect from running a single
TCP connection over any one of its available paths. A Multipath TCP TCP connection over any one of its available paths. A Multipath TCP
may not, however, be able to provide the same level of consistency of may not, however, be able to provide the same level of consistency of
throughput and latency as a single TCP connection. These, and other, throughput and latency as a single TCP connection. These, and other,
application considerations are discussed in detail in [8]. application considerations are discussed in detail in [8].
A multipath-capable equivalent of TCP MUST retain some level of A multipath-capable equivalent of TCP MUST retain some level of
backward compatibility with existing TCP APIs, so that existing backward compatibility with existing TCP APIs, so that existing
applications can use the newer transport merely by upgrading the applications can use the newer transport merely by upgrading the
operating systems of the end-hosts. This does not preclude the use operating systems of the end hosts. This does not preclude the use
of an advanced API to permit multipath-aware applications to specify of an advanced API to permit multipath-aware applications to specify
preferences, nor for users to configure their systems in a different preferences, nor for users to configure their systems in a different
way from the default, for example switching on or off the automatic way from the default, for example switching on or off the automatic
use of multipath extensions. use of multipath extensions.
It is possible for regular TCP sessions today to survive brief breaks It is possible for regular TCP sessions today to survive brief breaks
in connectivity by retaining state at end hosts before a timeout in connectivity by retaining state at end hosts before a timeout
occurs. It would be desirable to support similar session continuity occurs. It would be desirable to support similar session continuity
in MPTCP, however the circumstances could be different. Whilst in in MPTCP; however, the circumstances could be different. Whilst in
regular TCP the IP addresses will remain constant across the break in regular TCP the IP addresses will remain constant across the break in
connectivity, in MPTCP a different interface may appear. It is connectivity, in MPTCP a different interface may appear. It is
desirable (but not mandated) to support this kind of "break-before- desirable (but not mandated) to support this kind of "break-before-
make" session continuity. This places constraints on security make" session continuity. This places constraints on security
mechanisms, however, as discussed in Section 5.8. Timeouts for this mechanisms, however, as discussed in Section 5.8. Timeouts for this
function would be locally configured. function would be locally configured.
2.2.2. Network Compatibility 2.2.2. Network Compatibility
In the traditional Internet architecture, network devices operate at In the traditional Internet architecture, network devices operate at
the network layer and lower layers, with the layers above the network the network layer and lower layers, with the layers above the network
layer instantiated only at the end-hosts. While this architecture, layer instantiated only at the end hosts. While this architecture,
shown in Figure 2, was initially largely adhered to, this layering no shown in Figure 2, was initially largely adhered to, this layering no
longer reflects the "ground truth" in the Internet with the longer reflects the "ground truth" in the Internet with the
proliferation of middleboxes [9]. Middleboxes routinely interpose on proliferation of middleboxes [9]. Middleboxes routinely interpose on
the transport layer; sometimes even completely terminating transport the transport layer; sometimes even completely terminating transport
connections, thus leaving the application layer as the first real connections, thus leaving the application layer as the first real
end-to-end layer, as shown in Figure 3. end-to-end layer, as shown in Figure 3.
+-------------+ +-------------+ +-------------+ +-------------+
| Application |<------------ end-to-end ------------->| Application | | Application |<------------ end-to-end ------------->| Application |
+-------------+ +-------------+ +-------------+ +-------------+
skipping to change at page 9, line 11 skipping to change at page 9, line 23
End Host Router NAT, or Proxy End Host End Host Router NAT, or Proxy End Host
Figure 3: Internet Reality Figure 3: Internet Reality
Middleboxes that interpose on the transport layer result in loss of Middleboxes that interpose on the transport layer result in loss of
"fate-sharing" [10], that is, they often hold "hard" state that, when "fate-sharing" [10], that is, they often hold "hard" state that, when
lost or corrupted, results in loss or corruption of the end-to-end lost or corrupted, results in loss or corruption of the end-to-end
transport connection. transport connection.
The network compatibility goal requires that the multipath extension The network compatibility goal requires that the multipath extension
to TCP retains compatibility with the Internet as it exists today, to TCP retain compatibility with the Internet as it exists today,
including making reasonable efforts to be able to traverse including making reasonable efforts to be able to traverse
predominant middleboxes such as firewalls, NATs, and performance predominant middleboxes such as firewalls, NATs, and performance-
enhancing proxies [9]. This requirement comes from recognizing enhancing proxies [9]. This requirement comes from recognizing
middleboxes as a significant deployment bottleneck for any transport middleboxes as a significant deployment bottleneck for any transport
that is not TCP or UDP, and constrains Multipath TCP to appear as TCP that is not TCP or UDP, and constrains Multipath TCP to appear as TCP
does on the wire and to use established TCP extensions where does on the wire and to use established TCP extensions where
necessary. To ensure end-to-endness of the transport, we further necessary. To ensure "end-to-endness" of the transport, Multipath
require Multipath TCP to preserve fate-sharing without making any TCP MUST preserve fate-sharing without making any assumptions about
assumptions about middlebox behavior. middlebox behavior.
A detailed analysis of middlebox behaviour and the impact on the A detailed analysis of middlebox behavior and the impact on the
Multipath TCP architecture is presented in Section 7. In addition, Multipath TCP architecture is presented in Section 7. In addition,
network compatibility must be retained to the extent that Multipath network compatibility must be retained to the extent that Multipath
TCP MUST fall back to regular TCP if there are insurmountable TCP MUST fall back to regular TCP if there are insurmountable
incompatibilities for the multipath extension on a path. incompatibilities for the multipath extension on a path.
Middleboxes may also cause some TCP features to be able to exist on Middleboxes may also cause some TCP features to be able to exist on
one subflow but not another. Typically these will be at the subflow one subflow but not another. Typically, these will be at the subflow
level (such as SACK [11]) and thus do not affect the connection-level level (such as selective acknowledgment (SACK) [11]) and thus do not
behaviour. In the future, any proposed TCP connection-level affect the connection-level behavior. In the future, any proposed
extensions should consider how they can co-exist with MPTCP. TCP connection-level extensions should consider how they can coexist
with MPTCP.
The modifications to support Multipath TCP remain at the transport The modifications to support Multipath TCP remain at the transport
layer, although some knowledge of the underlying network layer is layer, although some knowledge of the underlying network layer is
required. Multipath TCP SHOULD work with IPv4 and IPv6 required. Multipath TCP SHOULD work with IPv4 and IPv6
interchangeably, i.e. one connection may operate over both IPv4 and interchangeably, i.e., one connection may operate over both IPv4 and
IPv6 networks. IPv6 networks.
2.2.3. Compatibility with other network users 2.2.3. Compatibility with Other Network Users
As a corollary to both network and application compatibility, the As a corollary to both network and application compatibility, the
architecture must enable new Multipath TCP flows to coexist architecture must enable new Multipath TCP flows to coexist
gracefully with existing single-path TCP flows, competing for gracefully with existing single-path TCP flows, competing for
bandwidth neither unduly aggressively nor unduly timidly (unless low- bandwidth neither unduly aggressively nor unduly timidly (unless low-
precedence operation is specifically requested by the application, precedence operation is specifically requested by the application,
such as with LEDBAT). The use of multiple paths MUST NOT unduly harm such as with LEDBAT). The use of multiple paths MUST NOT unduly harm
users using single-path TCP at shared bottlenecks, beyond the impact users using single-path TCP at shared bottlenecks, beyond the impact
that would occur from another single-path TCP flow. Multiple that would occur from another single-path TCP flow. Multiple
Multipath TCP flows on a shared bottleneck MUST share bandwidth Multipath TCP flows on a shared bottleneck MUST share bandwidth
between each other with similar fairness to that which occurs at a between each other with similar fairness to that which occurs at a
shared bottleneck with single-path TCP. shared bottleneck with single-path TCP.
2.3. Security Goals 2.3. Security Goals
The extension of TCP with multipath capabilities will bring with it a The extension of TCP with multipath capabilities will bring with it a
number of new threats, analysed in detail in [12]. The security goal number of new threats, analyzed in detail in [12]. The security goal
for Multipath TCP is to provide a service no less secure than for Multipath TCP is to provide a service no less secure than
regular, single-path TCP. This will be achieved through a regular, single-path TCP. This will be achieved through a
combination of existing TCP security mechanisms (potentially modified combination of existing TCP security mechanisms (potentially modified
to align with the Multipath TCP extensions) and of protection against to align with the Multipath TCP extensions) and of protection against
the new multipath threats identified. The design decisions derived the new multipath threats identified. The design decisions derived
from this goal are presented in Section 5.8. from this goal are presented in Section 5.8.
2.4. Related Protocols 2.4. Related Protocols
There are several similarities between SCTP [6] and MPTCP, in that There are several similarities between SCTP [6] and MPTCP, in that
both can make use of multiple addresses at end hosts to give some both can make use of multiple addresses at end hosts to give some
multi-path capability. In SCTP, the primary use case is to support multipath capability. In SCTP, the primary use case is to support
redundancy and mobility for multihomed hosts (i.e. a single path will redundancy and mobility for multihomed hosts (i.e., a single path
change one of its end host addresses); the simultaneous use of will change one of its end host addresses); the simultaneous use of
multiple paths is not supported . Extensions are proposed to support multiple paths is not supported. Extensions are proposed to support
simultaneous multipath transport [13], but these are yet to be simultaneous multipath transport [13], but these are yet to be
standardised. By far the most widely used stream-based transport standardized. By far the most widely used stream-based transport
protocol is, however, TCP [1], and SCTP does not meet the network and protocol is, however, TCP [1], and SCTP does not meet the network and
application compatibility goals specified in Section 2.2. For application compatibility goals specified in Section 2.2. For
network compatibility, there are issues with various middleboxes network compatibility, there are issues with various middleboxes
(especially NATs) that are unaware of SCTP and consequently end up (especially NATs) that are unaware of SCTP and consequently end up
blocking it. For application compatibility, applications need to blocking it. For application compatibility, applications need to
actively choose to use SCTP, and with the deployment issues very few actively choose to use SCTP, and with the deployment issues, very few
choose to do so. MPTCP's compatibility goals are in part based on choose to do so. MPTCP's compatibility goals are in part based on
these observations of SCTP's deployment issues. these observations of SCTP's deployment issues.
3. An Architectural Basis For Multipath TCP 3. An Architectural Basis for Multipath TCP
We now present one possible transport architecture that we believe This section presents one possible transport architecture that the
can effectively support the goals for Multipath TCP. The new authors believe can effectively support the goals for Multipath TCP.
Internet model described here is based on ideas proposed earlier in The new Internet model described here is based on ideas proposed
Tng ("Transport next-generation") [14]. While by no means the only earlier in Tng ("Transport next-generation") [14]. While by no means
possible architecture supporting multipath transport, Tng the only possible architecture supporting multipath transport, Tng
incorporates many lessons learned from previous transport research incorporates many lessons learned from previous transport research
and development practice, and offers a strong starting point from and development practice, and offers a strong starting point from
which to consider the extant Internet architecture and its bearing on which to consider the extant Internet architecture and its bearing on
the design of any new Internet transports or transport extensions. the design of any new Internet transports or transport extensions.
+------------------+ +------------------+
| Application | | Application |
+------------------+ ^ Application-oriented transport +------------------+ ^ Application-oriented transport
| | | functions (Semantic Layer) | | | functions (Semantic Layer)
+ - - Transport - -+ ---------------------------------- + - - Transport - -+ ----------------------------------
skipping to change at page 11, line 47 skipping to change at page 12, line 17
+-------------+ +-------------+ +-------------+ +-------------+
| Semantic |<------------ end-to-end ------------->| Semantic | | Semantic |<------------ end-to-end ------------->| Semantic |
+-------------+ +-------------+ +-------------+ +-------------+ +-------------+ +-------------+ +-------------+ +-------------+
|Flow+Endpoint|<->|Flow+Endpoint|<->|Flow+Endpoint|<->|Flow+Endpoint| |Flow+Endpoint|<->|Flow+Endpoint|<->|Flow+Endpoint|<->|Flow+Endpoint|
+-------------+ +-------------+ +-------------+ +-------------+ +-------------+ +-------------+ +-------------+ +-------------+
| Network |<->| Network |<->| Network |<->| Network | | Network |<->| Network |<->| Network |<->| Network |
+-------------+ +-------------+ +-------------+ +-------------+ +-------------+ +-------------+ +-------------+ +-------------+
Firewall Performance Firewall Performance
End Host or NAT Enhancing Proxy End Host End Host or NAT Enhancing Proxy End Host
Figure 5: Middleboxes in the new Internet model Figure 5: Middleboxes in the New Internet Model
MPTCP's architectural design follows Tng's decomposition as shown in MPTCP's architectural design follows Tng's decomposition as shown in
Figure 6. MPTCP, which provides application compatibility through Figure 6. MPTCP, which provides application compatibility through
the preservation of TCP-like semantics of global ordering of the preservation of TCP-like semantics of global ordering of
application data and reliability, is an instantiation of the application data and reliability, is an instantiation of the
"application-oriented" Semantic layer; whereas the subflow TCP "application-oriented" Semantic layer; whereas the subflow TCP
component, which provides network compatibility by appearing and component, which provides network compatibility by appearing and
behaving as a TCP flow in the network, is an instantiation of the behaving as a TCP flow in the network, is an instantiation of the
"network-oriented" Flow+Endpoint layer. "network-oriented" Flow+Endpoint layer.
+--------------------------+ +-------------------------------+ +--------------------------+ +-------------------------------+
| Application | | Application | | Application | | Application |
+--------------------------+ +-------------------------------+ +--------------------------+ +-------------------------------+
| Semantic | | MPTCP | | Semantic | | MPTCP |
|------------+-------------| + - - - - - - - + - - - - - - - + |------------+-------------| + - - - - - - - + - - - - - - - +
| Flow+Endpt | Flow+Endpt | | Subflow (TCP) | Subflow (TCP) | | Flow+Endpt | Flow+Endpt | | Subflow (TCP) | Subflow (TCP) |
+------------+-------------+ +---------------+---------------+ +------------+-------------+ +---------------+---------------+
| Network | Network | | IP | IP | | Network | Network | | IP | IP |
+------------+-------------+ +---------------+---------------+ +------------+-------------+ +---------------+---------------+
Figure 6: Relationship between Tng (left) and MPTCP (right) Figure 6: Relationship between Tng (Left) and MPTCP (Right)
As a protocol extension to TCP, MPTCP thus explicitly acknowledges As a protocol extension to TCP, MPTCP thus explicitly acknowledges
middleboxes in its design, and specifies a protocol that operates at middleboxes in its design, and specifies a protocol that operates at
two scales: the MPTCP component operates end-to-end, while it allows two scales: the MPTCP component operates end-to-end, while it allows
the TCP component to operate segment-by-segment. the TCP component to operate segment-by-segment.
4. A Functional Decomposition of MPTCP 4. A Functional Decomposition of MPTCP
The previous two sections have discussed the goals for a Multipath The previous two sections have discussed the goals for a Multipath
TCP design, and provided a basis for decomposing the functions of a TCP design, and provided a basis for decomposing the functions of a
skipping to change at page 13, line 29 skipping to change at page 13, line 37
implement the following functions: implement the following functions:
o Path Management: This is the function to detect and use multiple o Path Management: This is the function to detect and use multiple
paths between two hosts. MPTCP uses the presence of multiple IP paths between two hosts. MPTCP uses the presence of multiple IP
addresses at one or both of the hosts as an indicator of this. addresses at one or both of the hosts as an indicator of this.
The path management features of the MPTCP protocol are the The path management features of the MPTCP protocol are the
mechanisms to signal alternative addresses to hosts, and mechanisms to signal alternative addresses to hosts, and
mechanisms to set up new subflows joined to an existing MPTCP mechanisms to set up new subflows joined to an existing MPTCP
connection. connection.
o Packet Scheduling: This function breaks the bytestream received o Packet Scheduling: This function breaks the byte stream received
from the application into segments to be transmitted on one of the from the application into segments to be transmitted on one of the
available subflows. The MPTCP design makes use of a data sequence available subflows. The MPTCP design makes use of a data sequence
mapping, associating segments sent on different subflows to a mapping, associating segments sent on different subflows to a
connection-level sequence numbering, thus allowing segments sent connection-level sequence numbering, thus allowing segments sent
on different subflows to be correctly re-ordered at the receiver. on different subflows to be correctly re-ordered at the receiver.
The packet scheduler is dependent upon information about the The packet scheduler is dependent upon information about the
availability of paths exposed by the path management component, availability of paths exposed by the path management component,
and then makes use of the subflows to transmit queued segments. and then makes use of the subflows to transmit queued segments.
This function is also responsible for connection-level re-ordering This function is also responsible for connection-level re-ordering
on receipt of packets from the TCP subflows, according to the on receipt of packets from the TCP subflows, according to the
skipping to change at page 13, line 45 skipping to change at page 14, line 4
The packet scheduler is dependent upon information about the The packet scheduler is dependent upon information about the
availability of paths exposed by the path management component, availability of paths exposed by the path management component,
and then makes use of the subflows to transmit queued segments. and then makes use of the subflows to transmit queued segments.
This function is also responsible for connection-level re-ordering This function is also responsible for connection-level re-ordering
on receipt of packets from the TCP subflows, according to the on receipt of packets from the TCP subflows, according to the
attached data sequence mappings. attached data sequence mappings.
o Subflow (single-path TCP) Interface: A subflow component takes o Subflow (single-path TCP) Interface: A subflow component takes
segments from the packet-scheduling component and transmits them segments from the packet-scheduling component and transmits them
over the specified path, ensuring detectable delivery to the host. over the specified path, ensuring detectable delivery to the host.
MPTCP uses TCP underneath for network compatibility; TCP ensures MPTCP uses TCP underneath for network compatibility; TCP ensures
in-order, reliable delivery. TCP adds its own sequence numbers to in-order, reliable delivery. TCP adds its own sequence numbers to
the segments; these are used to detect and retransmit lost packets the segments; these are used to detect and retransmit lost packets
at the subflow layer. On receipt, the subflow passes its at the subflow layer. On receipt, the subflow passes its
reassembled data to the packet scheduling component for reassembled data to the packet scheduling component for
connection-level reassembly; the data sequence mapping from the connection-level reassembly; the data sequence mapping from the
sender's packet scheduling component allows re-ordering of the sender's packet scheduling component allows re-ordering of the
entire bytestream. entire byte stream.
o Congestion Control: This function coordinates congestion control o Congestion Control: This function coordinates congestion control
across the subflows. As specified, this congestion control across the subflows. As specified, this congestion control
algorithm MUST ensure that a MPTCP connection does not unfairly algorithm MUST ensure that an MPTCP connection does not unfairly
take more bandwidth than a single path TCP flow would take at a take more bandwidth than a single path TCP flow would take at a
shared bottleneck. An algorithm to support this is specified in shared bottleneck. An algorithm to support this is specified in
[7]. [7].
These functions fit together as follows. The Path Management looks These functions fit together as follows. The path management looks
after the discovery (and if necessary, initialisation) of multiple after the discovery (and if necessary, initialization) of multiple
paths between two hosts. The Packet Scheduler then receives a stream paths between two hosts. The packet scheduler then receives a stream
of data from the application destined for the network, and undertakes of data from the application destined for the network, and undertakes
the necessary operations on it (such as segmenting the data into the necessary operations on it (such as segmenting the data into
connection-level segments, and adding a connection-level sequence connection-level segments, and adding a connection-level sequence
number) before sending it on to a subflow. The subflow then adds its number) before sending it on to a subflow. The subflow then adds its
own sequence number, ACKs, and passes them to network. The receiving own sequence number, ACKs, and passes them to network. The receiving
subflow re-orders data (if necessary) and passes it to the packet subflow re-orders data (if necessary) and passes it to the packet
scheduling component, which performs connection level re-ordering, scheduling component, which performs connection level re-ordering,
and sends the data stream to the application. Finally, the and sends the data stream to the application. Finally, the
congestion control component exists as part of the packet scheduling, congestion control component exists as part of the packet scheduling,
in order to schedule which segments should be sent at what rate on in order to schedule which segments should be sent at what rate on
which subflow. which subflow.
5. High-Level Design Decisions 5. High-Level Design Decisions
There is seemingly a wide range of choices when designing a multipath There is seemingly a wide range of choices when designing a multipath
extension to TCP. However, the goals as discussed earlier in this extension to TCP. However, the goals as discussed earlier in this
document constrain the possible solutions, leaving relative little document constrain the possible solutions, leaving relative little
choice in many areas. Here, we outline high-level design choices choice in many areas. This section outlines high-level design
that draw from the architectural basis discussed earlier in choices that draw from the architectural basis discussed earlier in
Section 3, which the design of MPTCP [5] takes into account. Section 3, which the design of MPTCP [5] takes into account.
5.1. Sequence Numbering 5.1. Sequence Numbering
MPTCP uses two levels of sequence spaces: a connection level sequence MPTCP uses two levels of sequence spaces: a connection-level sequence
number, and another sequence number for each subflow. This permits number and another sequence number for each subflow. This permits
connection-level segmentation and reassembly, and retransmission of connection-level segmentation and reassembly and retransmission of
the same part of connection-level sequence space on different the same part of connection-level sequence space on different
subflow-level sequence space. subflow-level sequence space.
The alternative approach would be to use a single connection level The alternative approach would be to use a single connection-level
sequence number, which gets sent on multiple subflows. This has two sequence number, which gets sent on multiple subflows. This has two
problems: first, the individual subflows will appear to the network problems: first, the individual subflows will appear to the network
as TCP sessions with gaps in the sequence space; this in turn may as TCP sessions with gaps in the sequence space; this in turn may
upset certain middleboxes such as intrusion detection systems, or upset certain middleboxes such as intrusion detection systems, or
certain transparent proxies, and would thus go against the network certain transparent proxies, and would thus go against the network
compatibility goal. Second, the sender would not be able to compatibility goal. Second, the sender would not be able to
attribute packet losses or receptions to the correct path when the attribute packet losses or receptions to the correct path when the
same segment is sent on multiple paths (i.e. in the case of same segment is sent on multiple paths (i.e., in the case of
retransmissions). retransmissions).
The sender must be able to tell the receiver how to reassemble the The sender must be able to tell the receiver how to reassemble the
data, for delivery to the application. In order to achieve this, the data, for delivery to the application. In order to achieve this, the
receiver must determine how subflow-level data (carrying subflow receiver must determine how subflow-level data (carrying subflow
sequence numbers) maps at the connection level. We refer to this as sequence numbers) maps at the connection level. This is referred to
the Data Sequence Mapping. This mapping takes the form (data seq, as the "data sequence mapping". This mapping can be represented as a
subflow seq, length), i.e. for a given number of bytes (the length), tuple of (data sequence number, subflow sequence number, length),
the subflow sequence space beginning at the given sequence number i.e., for a given number of bytes (the length), the subflow sequence
maps to the connection-level sequence space (beginning at the given space beginning at the given sequence number maps to the connection-
data seq number). This information could conceivably have various level sequence space (beginning at the given data sequence number).
sources. This information could conceivably have various sources.
One option to signal the Data Sequence Mapping would be to use One option to signal the data sequence mapping would be to use
existing fields in the TCP segment (such as subflow seqno, length) existing fields in the TCP segment (such as subflow sequence number,
and only add the data sequence number to each segment, for instance length) and add only the data sequence number to each segment, for
as a TCP option. This would be vulnerable, however, to middleboxes instance, as a TCP option. This would be vulnerable, however, to
that resegment or assemble data, since there is no specified middleboxes that re-segment or assemble data, since there is no
behaviour for coalescing TCP options. If one signalled (data seqno, specified behavior for coalescing TCP options. If one signaled (data
length), this would still be vulnerable to middleboxes that coalesce sequence number, length), this would still be vulnerable to
segments and do not understand MPTCP signalling so do not correctly middleboxes that coalesce segments and do not understand MPTCP
rewrite the options. signaling so do not correctly rewrite the options.
Because of these potential issues, the design decision taken in the Because of these potential issues, the design decision taken in the
MPTCP protocol is that whenever a mapping for subflow data needs to MPTCP protocol is that whenever a mapping for subflow data needs to
be conveyed to the other host, all three pieces of data (data seq, be conveyed to the other host, all three pieces of data (data seq,
subflow seq, length) must be sent. To reduce the overhead, it would subflow seq, length) must be sent. To reduce the overhead, it would
be permissible for the mapping to be sent periodically and cover more be permissible for the mapping to be sent periodically and cover more
than a single segment. Further experimentation is required to than a single segment. Further experimentation is required to
determine what tradeoffs exist regarding the frequency at which determine what tradeoffs exist regarding the frequency at which
mappings should be sent. It could also be excluded entirely in the mappings should be sent. It could also be excluded entirely in the
case of a connection before more than one subflow is used, where the case of a connection before more than one subflow is used, where the
data-level and subflow-level sequence space is the same. data-level and subflow-level sequence space is the same.
5.2. Reliability and Retransmissions 5.2. Reliability and Retransmissions
MPTCP features acknowledgements at connection-level as well as MPTCP features acknowledgements at connection-level as well as
subflow-level acknowledgements, in order to provide a robust service subflow-level acknowledgements, in order to provide a robust service
to the application. to the application.
Under normal behaviour, MPTCP can use the data sequence mapping and Under normal behavior, MPTCP could use the data sequence mapping and
subflow ACKs to decide when a connection-level segment was received. subflow ACKs to decide when a connection-level segment was received.
The transmission of TCP ACKs for a subflow are handled entirely at The transmission of TCP ACKs for a subflow are handled entirely at
the subflow level, in order to maintain TCP semantics and trigger the subflow level, in order to maintain TCP semantics and trigger
subflow-level retransmissions. This has certain implications on end- subflow-level retransmissions. This has certain implications on end-
to-end semantics. It means that once a segment is ACKed at the to-end semantics. It would mean that once a segment is ACKed at the
subflow level it cannot be discarded in the re-order buffer at the subflow level, it cannot be discarded in the re-order buffer at the
connection level. Secondly, unlike in standard TCP, a receiver connection level. Secondly, unlike in standard TCP, a receiver
cannot simply drop out-of-order segments if needed (for instance, due cannot simply drop out-of-order segments if needed (for instance, due
to memory pressure). Under certain circumstances, therefore, it may to memory pressure). Under certain circumstances, it may be
be desirable to drop segments after acknowledgement on the subflow desirable to drop segments after acknowledgement on the subflow but
but before delivery to the application, and this can be facilitated before delivery to the application, and this can be facilitated by a
by a connection-level acknowledgement. connection-level acknowledgement.
Furthermore, it is possible to conceive of some cases where Furthermore, it is possible to conceive of some cases where
connection-level acknowledgements could improve robustness. Consider connection-level acknowledgements could improve robustness. Consider
a subflow traversing a transparent proxy: if the proxy ACKs a segment a subflow traversing a transparent proxy: if the proxy ACKs a segment
and then crashes, the sender will not retransmit the lost segment on and then crashes, the sender will not retransmit the lost segment on
another subflow, as it thinks the segment has been received. The another subflow, as it thinks the segment has been received. The
connection grinds to a halt despite having other working subflows, connection grinds to a halt despite having other working subflows,
and the sender would be unable to determine the cause of the problem. and the sender would be unable to determine the cause of the problem.
An example situation where this may occur would be mobility between An example situation where this may occur would be mobility between
wireless access points, each of which operates a transport-level wireless access points, each of which operates a transport-level
proxy. Finally, as an optimisation, it may be feasible for a proxy. Finally, as an optimization, it may be feasible for a
connection-level acknowledgement to be transmitted over the shortest connection-level acknowledgement to be transmitted over the shortest
Round-Trip Time (RTT) path, potentially reducing send buffer Round-Trip Time (RTT) path, potentially reducing send buffer
requirements (see Section 5.3). requirements (see Section 5.3).
Therefore, to provide a fully robust multipath TCP solution given the Therefore, to provide a fully robust multipath TCP solution given the
above constraints, MPTCP for use on the public Internet MUST feature above constraints, MPTCP for use on the public Internet MUST feature
explicit connection-level acknowledgements, in addition to subflow- explicit connection-level acknowledgements, in addition to subflow-
level acknowledgements. A connection-level acknowledgement would level acknowledgements. A connection-level acknowledgement would
only be required in order to signal when the receive window moves only be required in order to signal when the receive window moves
forward; the heuristics for using such a signal are discussed in more forward; the heuristics for using such a signal are discussed in more
detail in the protocol specification [5]. detail in the protocol specification [5].
Regarding retransmissions, it MUST be possible for a segments to be Regarding retransmissions, it MUST be possible for a segment to be
retransmitted on a different subflow to that on which it was retransmitted on a different subflow from that on which it was
originally sent. This is one of MPTCP's core goals, in order to originally sent. This is one of MPTCP's core goals, in order to
maintain integrity during temporary or permanent subflow failure, and maintain integrity during temporary or permanent subflow failure, and
this is enabled by the dual sequence number space. this is enabled by the dual sequence number space.
The scheduling of retransmissions will have significant impact on The scheduling of retransmissions will have significant impact on
MPTCP user experience. The current MPTCP specification suggests that MPTCP user experience. The current MPTCP specification suggests that
data outstanding on subflows that have timed out should be data outstanding on subflows that have timed out should be
rescheduled for transmission on different subflows. This behaviour rescheduled for transmission on different subflows. This behavior
aims to minimize disruption when a path breaks, and uses the first aims to minimize disruption when a path breaks, and uses the first
timeout as indicators. More conservative versions would be to use timeout as indicators. More conservative versions would be to use
second or third timeouts for the same segment. second or third timeouts for the same segment.
Typically, fast retransmit on an individual subflow will not trigger Typically, fast retransmit on an individual subflow will not trigger
retransmission on another subflow, although this may still be retransmission on another subflow, although this may still be
desirable in certain cases, for instance to reduce the receive buffer desirable in certain cases, for instance, to reduce the receive
requirements. However, in all cases with retransmissions on buffer requirements. However, in all cases with retransmissions on
different subflows, the lost segments SHOULD still be sent on the different subflows, the lost segments SHOULD still be sent on the
path that lost them. This is currently believed to be necessary to path that lost them. This is currently believed to be necessary to
maintain subflow integrity, as per the network compatibility goal. maintain subflow integrity, as per the network compatibility goal.
By doing this, some efficiency is lost, and it is unclear at this By doing this, some efficiency is lost, and it is unclear at this
point what the optimal retransmit strategy is. point what the optimal retransmit strategy is.
Large-scale experiments are therefore required in order to determine Large-scale experiments are therefore required in order to determine
the most appropriate retransmission strategy, and recommendations the most appropriate retransmission strategy, and recommendations
will be refined once more information is available. will be refined once more information is available.
5.3. Buffers 5.3. Buffers
To ensure in-order delivery, MPTCP must use a connection level To ensure in-order delivery, MPTCP must use a connection level
receive buffer, where segments are placed until they are in order and receive buffer, where segments are placed until they are in order and
can be read by the application. can be read by the application.
In regular, single-path TCP, it is usually recommended to set the In regular, single-path TCP, it is usually recommended to set the
receive buffer to 2*BDP (Bandwidth-Delay Product, i.e. BDP = BW*RTT, receive buffer to 2*BDP (Bandwidth-Delay Product, i.e., BDP = BW*RTT,
where BW = Bandwidth and RTT = Round-Trip Time). One BDP allows where BW = Bandwidth and RTT = Round-Trip Time). One BDP allows
supporting reordering of segments by the network. The other BDP supporting reordering of segments by the network. The other BDP
allows the connection to continue during fast retransmit: when a allows the connection to continue during fast retransmit: when a
segment is fast retransmitted, the receiver must be able to store segment is fast retransmitted, the receiver must be able to store
incoming data during one more RTT. incoming data during one more RTT.
For MPTCP, the story is a bit more complicated. The ultimate goal is For MPTCP, the story is a bit more complicated. The ultimate goal is
that a subflow packet loss or subflow failure should not affect the that a subflow packet loss or subflow failure should not affect the
throughput of other working subflows; the receiver should have enough throughput of other working subflows; the receiver should have enough
buffering to store all data until the missing segment is re- buffering to store all data until the missing segment is re-
transmitted and reaches the destination. transmitted and reaches the destination.
The worst case scenario would be when the subflow with the highest The worst-case scenario would be when the subflow with the highest
RTT/RTO (Round-Trip Time or Retransmission TimeOut) experiences a RTT/RTO (Round-Trip Time or Retransmission TimeOut) experiences a
timeout; in that case the receiver has to buffer data from all timeout; in that case, the receiver has to buffer data from all
subflows for the duration of the RTO. Thus, the smallest connection- subflows for the duration of the RTO. Thus, the smallest connection-
level receive buffer that would be needed to avoid stalling with level receive buffer that would be needed to avoid stalling with
subflow failures is sum(BW_i)*RTO_max, where BW_i = Bandwidth for subflow failures is sum(BW_i)*RTO_max, where BW_i = Bandwidth for
each subflow and RTO_max is the largest RTO across all subflows. each subflow and RTO_max is the largest RTO across all subflows.
This is an order of magnitude more than the receive buffer required This is an order of magnitude more than the receive buffer required
for a single connection, and is probably too expensive for practical for a single connection, and is probably too expensive for practical
purposes. A more sensible requirement is to avoid stalls in the purposes. A more sensible requirement is to avoid stalls in the
absence of timeouts. Therefore, the RECOMMENDED receive buffer is absence of timeouts. Therefore, the RECOMMENDED receive buffer is
2*sum(BW_i)*RTT_max, where RTT_max is the largest RTT across all 2*sum(BW_i)*RTT_max, where RTT_max is the largest RTT across all
subflows. This buffer sizing ensures subflows do not stall when fast subflows. This buffer sizing ensures subflows do not stall when fast
retransmit is triggered on any subflow. retransmit is triggered on any subflow.
The resulting buffer size should be small enough for practical use. The resulting buffer size should be small enough for practical use.
However, there may be extreme cases where fast, high throughput paths However, there may be extreme cases where fast, high throughput paths
(e.g. 100Mb/s, 10ms RTT) are used in conjunction with slow paths (e.g., 100 Mb/s, 10 ms RTT) are used in conjunction with slow paths
(e.g. 1Mb/s, 1000ms RTT). In that case the required receive buffer (e.g., 1 Mb/s, 1000 ms RTT). In that case, the required receive
would be 12.5MB, which is likely too big. In extreme cases such as buffer would be 12.5 MB, which is likely too big. In extreme cases
this example, it may be prudent to only use some of the fastest such as this example, it may be prudent to only use some of the
available paths for the MPTCP connection, potentially using the slow fastest available paths for the MPTCP connection, potentially using
path(s) for backup only. the slow path(s) for backup only.
Send Buffer: The RECOMMENDED send buffer is the same size as the Send Buffer: The RECOMMENDED send buffer is the same size as the
recommended receive buffer i.e., 2*sum(BW_i)*RTT_max. This is recommended receive buffer, i.e., 2*sum(BW_i)*RTT_max. This is
because the sender must store locally the segments sent but because the sender must locally store the segments sent but
unacknowledged by the connection level ACK. The send buffer size unacknowledged by the connection level ACK. The send buffer size
matters particularly for hosts that maintain a large number of matters particularly for hosts that maintain a large number of
ongoing connections. If the required send buffer is too large, a ongoing connections. If the required send buffer is too large, a
host can choose to only send data on the fast subflows, using the host can choose to only send data on the fast subflows, using the
slow subflows only in cases of failure. slow subflows only in cases of failure.
5.4. Signalling 5.4. Signaling
Since MPTCP uses TCP as its subflow transport mechanism, a MPTCP Since MPTCP uses TCP as its subflow transport mechanism, an MPTCP
connection will also begin as a single TCP connection. Nevertheless, connection will also begin as a single TCP connection. Nevertheless,
it must signal to the peer that it supports MPTCP and wishes to use it must signal to the peer that it supports MPTCP and wishes to use
it on this connection. As such, a TCP Option will be used to it on this connection. As such, a TCP option will be used to
transmit this information, since this is the established mechanism transmit this information, since this is the established mechanism
for indicating additional functionality on a TCP session. for indicating additional functionality on a TCP session.
In addition, further signalling is required during the operation of a In addition, further signaling is required during the operation of an
MPTCP session, such as that for reassembly for multiple subflows, and MPTCP session, such as that for reassembly across multiple subflows,
for informing the other host about potential other available and for informing the other host about other available IP addresses.
addresses.
The MPTCP protocol design will, however, use TCP Options for this The MPTCP protocol design will use TCP options for this additional
additional signalling. This has been chosen as the mechanism most signaling. This has been chosen as the mechanism most fitting in
fitting in with the goals as specified in Section 2. With this with the goals as specified in Section 2. With this mechanism, the
mechanism, the signalling requires to operate MPTCP is transported signaling required to operate MPTCP is transported separately from
separately from the data, allowing it to be created and processed the data, allowing it to be created and processed separately from the
separately from the data stream, and retaining architectural data stream, and retaining architectural compatibility with network
compatibility with network entities. entities.
This decision is the consensus of the Working Group (following This decision is the consensus of the Working Group (following
detailed discussions at IETF78), and the main reasons for this are as detailed discussions at IETF78), and the main reasons for this are as
follows: follows:
o TCP options are the traditional signalling method for TCP; o TCP options are the traditional signaling method for TCP;
o A TCP option on a SYN is the most compatible way for an end host o A TCP option on a SYN is the most compatible way for an end host
to signal it is MPTCP-capable; to signal it is MPTCP capable;
o If connection-level ACKs are signalled in the payload then they o If connection-level ACKs are signaled in the payload, then they
may suffer from packet loss and may be congestion-controlled, may suffer from packet loss and may be congestion-controlled,
which may affect the data throughput in the forward direction and which may affect the data throughput in the forward direction and
could lead to head-of-line blocking; could lead to head-of-line blocking;
o Middleboxes, such as NAT traversal helpers, can easily parse TCP o Middleboxes, such as NAT traversal helpers, can easily parse TCP
options, e. g., to rewrite addresses. options, e.g., to rewrite addresses.
On the other hand, the main drawbacks of TCP options compared to TLV On the other hand, the main drawbacks of TCP options compared to TLV
encoding in the payload are: encoding in the payload are the following:
o There is limited space for signalling messages; o There is limited space for signaling messages;
o A middlebox may, potentially, drop a packet with an unknown o A middlebox may, potentially, drop a packet with an unknown
option; option;
o The transport of control information in options is not necessarily o The transport of control information in options is not necessarily
reliable. reliable.
The detailed design of MPTCP alleviates these issues as far as The detailed design of MPTCP alleviates these issues as far as
possible by carefully considering the size of MPTCP options, and possible by carefully considering the size of MPTCP options and
seamlessly falling back to regular TCP on the loss of control data. seamlessly falling back to regular TCP on the loss of control data.
Both option and payload encoding may interfere with offloading of TCP Both option and payload encoding may interfere with offloading of TCP
processing to high speed network interface cards, such as processing to high-speed network interface cards, such as
segmentation, checksumming, and reassembly. For network cards segmentation, checksumming, and reassembly. For network cards
supporting MPTCP, signalling in TCP options should simplify supporting MPTCP, signaling in TCP options should simplify offloading
offloading due to the separate handling of MPTCP signalling and data. due to the separate handling of MPTCP signaling and data.
5.5. Path Management 5.5. Path Management
Currently, the network does not expose path diversity between pairs Currently, the network does not expose path diversity between pairs
of IP addresses. In order to achieve path diversity from today's IP of IP addresses. In order to achieve path diversity from today's IP
networks, in the typical case MPTCP uses multiple addresses at one or networks, in the typical case, MPTCP uses multiple addresses at one
both hosts to infer different paths across the network. It is or both hosts to infer different paths across the network. It is
expected that these paths, whilst not necessarily entirely non- expected that these paths, whilst not necessarily entirely non-
overlapping, will be sufficiently disjoint to allow multipath to overlapping, will be sufficiently disjoint to allow multipath to
achieve improved throughput and robustness. The use of multiple IP achieve improved throughput and robustness. The use of multiple IP
addresses is a simple mechanism that requires no additional features addresses is a simple mechanism that requires no additional features
in the network. in the network.
Multiple different (source, destination) address pairs will thus be Multiple different (source, destination) address pairs will thus be
used as path selectors in most cases. Each path will be identified used as path selectors in most cases. However, each path will be
by a standard five-tuple (i.e. source address, destination address, identified by a standard five-tuple (i.e., source address,
source port, destination port, protocol), however, which can allow destination address, source port, destination port, protocol), which
the extension of MPTCP to use ports as well as addresses as path can allow the extension of MPTCP to use ports as well as addresses as
selectors. This will allow hosts to use port-based load balancing path selectors. This will allow hosts to use port-based load
with MPTCP, for example if the network routes different ports over balancing with MPTCP, for example, if the network routes different
different paths (which may be the case with technologies such as ports over different paths (which may be the case with technologies
Equal Cost MultiPath (ECMP) routing [4]). It should be noted, such as Equal Cost MultiPath (ECMP) routing [4]). It should be
however, that ISPs often undertake traffic engineering in order to noted, however, that ISPs often undertake traffic engineering in
optimise resource utilisation within their networks, and care should order to optimize resource utilization within their networks, and
be taken (by both ISPs and developers) that MPTCP using broadly care should be taken (by both ISPs and developers) that MPTCP using
similar paths does not adversely interfere with this. broadly similar paths does not adversely interfere with this.
For increased chance of successfully setting up additional subflows For an increased chance of successfully setting up additional
(such as when one end is behind a firewall, NAT, or other restrictive subflows (such as when one end is behind a firewall, NAT, or other
middlebox), either host SHOULD be able to add new subflows to a MPTCP restrictive middlebox), either host SHOULD be able to add new
connection. MPTCP MUST be able to handle paths that appear and subflows to an MPTCP connection. MPTCP MUST be able to handle paths
disappear during the lifetime of a connection (for example, through that appear and disappear during the lifetime of a connection (for
the activation of an additional network interface). example, through the activation of an additional network interface).
The path management is a separate function from the packet The path management is a separate function from the packet
scheduling, subflow interface, and congestion control functions of scheduling, subflow interface, and congestion control functions of
MPTCP, as documented in Section 4. As such it would be feasible to MPTCP, as documented in Section 4. As such, it would be feasible to
replace this IP-address-based design with an alternative path replace this IP-address-based design with an alternative path
selection mechanism in the future, with no significant changes to the selection mechanism in the future, with no significant changes to the
other functional components. other functional components.
5.6. Connection Identification 5.6. Connection Identification
Since a MPTCP connection may not be bound to a traditional 5-tuple Since an MPTCP connection may not be bound to a traditional 5-tuple
(source address and port, destination address and port, protocol (source address and port, destination address and port, protocol
number) for the entirety of its existence, it is desirable to provide number) for the entirety of its existence, it is desirable to provide
a new mechanism for connection identification. This will be useful a new mechanism for connection identification. This will be useful
for MPTCP-aware applications, and for the MPTCP implementation (and for MPTCP-aware applications and for the MPTCP implementation (and
MPTCP-aware middleboxes) to have a unique identifier with which to MPTCP-aware middleboxes) to have a unique identifier with which to
associate the multiple subflows. associate the multiple subflows.
Therefore, each MPTCP connection requires a connection identifier at Therefore, each MPTCP connection requires a connection identifier at
each host, which is locally unique within that host. In many ways, each host, which is locally unique within that host. In many ways,
this is analogous to an ephemeral port number in regular TCP. The this is analogous to an ephemeral port number in regular TCP. The
manifestation and purpose of such an identifier is out of the scope manifestation and purpose of such an identifier is out of the scope
of this architecture document. of this architecture document.
Legacy applications will not, however, have access to this identifier Non-MPTCP-aware applications will not, however, have access to this
and in such cases a MPTCP connection will be identified by the identifier and in such cases an MPTCP connection will be identified
5-tuple of the first TCP subflow. It is out of the scope of this by the 5-tuple of the first TCP subflow. It is out of the scope of
document, however, to define the behaviour of the MPTCP this document, however, to define the behavior of the MPTCP
implementation if the first TCP subflow later fails. If there are implementation if the first TCP subflow later fails. If there are
MPTCP-unaware applications that make assumptions about continued MPTCP-unaware applications that make assumptions about continued
existence of the initial address pair, their behaviour could be existence of the initial address pair, their behavior could be
disrupted by carrying on regardless. It is expected that this is a disrupted by carrying on regardless. It is expected that this is a
very small, possibly negligible, set of applications, however. MPTCP very small, possibly negligible, set of applications, however. MPTCP
MUST NOT be used for applications that request to bind to a specific MUST NOT be used for applications that request to bind to a specific
address or interface, since such applications are making a deliberate address or interface, since such applications are making a deliberate
choice of path in use. choice of path in use.
Since the requirements of applications are not clear at this stage, Since the requirements of applications are not clear at this stage,
however, it is as yet unconfirmed whether carrying on in the event of however, it is as yet unconfirmed whether carrying on in the event of
the loss of the initial address pair would be a damaging assumption the loss of the initial address pair would be a damaging assumption
to make. This behaviour will be an implementation-specific solution, to make. This behavior will be an implementation-specific solution,
and as such it is expected to be chosen by implementors once more and as such it is expected to be chosen by implementors once more
research has been undertaken to determine its impact. research has been undertaken to determine its impact.
5.7. Congestion Control 5.7. Congestion Control
As discussed in network-layer compatibility requirements As discussed in network-layer compatibility requirements
Section 2.2.3, there are three goals for the congestion control Section 2.2.3, there are three goals for the congestion control
algorithms used by a MPTCP implementation: improve throughput (at algorithms used by an MPTCP implementation: improve throughput (at
least as well as a single-path TCP connection would perform); do no least as well as a single-path TCP connection would perform); do no
harm to other network users (do not take up more capacity on any one harm to other network users (do not take up more capacity on any one
path than if it was a single path flow using only that route - this path than if it was a single path flow using only that route -- this
is particularly relevant for shared bottlenecks); and balance is particularly relevant for shared bottlenecks); and balance
congestion by moving traffic away from the most congested paths. To congestion by moving traffic away from the most congested paths. To
achieve these goals, the congestion control algorithms on each achieve these goals, the congestion control algorithms on each
subflow must be coupled in some way. A proposal for a suitable subflow must be coupled in some way. A proposal for a suitable
congestion control algorithm is given in [7]. congestion control algorithm is given in [7].
5.8. Security 5.8. Security
A detailed threat analysis for Multipath TCP is presented in a A detailed threat analysis for Multipath TCP is presented in a
separate document [12]. This focuses on flooding attacks and separate document [12]. That document focuses on flooding attacks
hijacking attacks that can be launched against a Multipath TCP and hijacking attacks that can be launched against a Multipath TCP
connection. connection.
The basic security goal of Multipath TCP, as introduced in The basic security goal of Multipath TCP, as introduced in
Section 2.3, can be stated as: "provide a solution that is no worse Section 2.3, can be stated as: "provide a solution that is no worse
than standard TCP". than standard TCP".
From the threat analysis, and with this goal in mind, three key From the threat analysis, and with this goal in mind, three key
security requirements can be identified. A multi-addressed Multipath security requirements can be identified. A multi-addressed Multipath
TCP SHOULD be able to: TCP SHOULD be able to do the following:
o Provide a mechanism to confirm that the parties in a subflow o Provide a mechanism to confirm that the parties in a subflow
handshake are the same as in the original connection setup (e.g. handshake are the same as in the original connection setup (e.g.,
require use of a key exchanged in the initial handshake in the require use of a key exchanged in the initial handshake in the
subflow handshake, to limit the scope for hijacking attacks). subflow handshake, to limit the scope for hijacking attacks).
o Provide verification that the peer can receive traffic at a new o Provide verification that the peer can receive traffic at a new
address before adding it (i.e. verify that the address belongs to address before adding it (i.e., verify that the address belongs to
the other host, to prevent flooding attacks). the other host, to prevent flooding attacks).
o Provide replay protection, i.e. ensure that a request to add/ o Provide replay protection, i.e., ensure that a request to add/
remove a subflow is 'fresh'. remove a subflow is 'fresh'.
Additional mechanisms have been deployed as part of standard TCP Additional mechanisms have been deployed as part of standard TCP
stacks to provide resistance to Denial-of-Service attacks. For stacks to provide resistance to Denial-of-Service (DoS) attacks. For
example, there are various mechanisms to protect against TCP reset example, there are various mechanisms to protect against TCP reset
attacks [18], and Multipath TCP should continue to support similar attacks [18], and Multipath TCP should continue to support similar
protection. In addition, TCP SYN Cookies [19] were developed to protection. In addition, TCP SYN Cookies [19] were developed to
allow a TCP server to defer the creation of session state in the allow a TCP server to defer the creation of session state in the
SYN_RCVD state, and remain stateless until the ESTABLISHED state had SYN_RCVD state, and remain stateless until the ESTABLISHED state had
been reached. Multipath TCP should, ideally, continue to provide been reached. Multipath TCP should, ideally, continue to provide
such functionality and, at a minimum, avoid significant computational such functionality and, at a minimum, avoid significant computational
burden prior to reaching the ESTABLISHED state (of the Multipath TCP burden prior to reaching the ESTABLISHED state (of the Multipath TCP
connection as a whole). connection as a whole).
skipping to change at page 22, line 42 skipping to change at page 23, line 12
in mind, and the protocol specification [5] will document how these in mind, and the protocol specification [5] will document how these
are met. are met.
6. Software Interactions 6. Software Interactions
6.1. Interactions with Applications 6.1. Interactions with Applications
In the case of applications that have used an existing API call to In the case of applications that have used an existing API call to
bind to a specific address or interface, the MPTCP extension MUST NOT bind to a specific address or interface, the MPTCP extension MUST NOT
be used. This is because the applications are indicating a clear be used. This is because the applications are indicating a clear
choice of path to use and thus will have expectations of behaviour choice of path to use and thus will have expectations of behavior
that must be maintained, in order to adhere to the application that must be maintained, in order to adhere to the application
compatibility goals. compatibility goals.
Interactions with applications are presented in [8] - including, but Interactions with applications are presented in [8] -- including, but
not limited to, performances changes that may be expected, semantic not limited to, performances changes that may be expected, semantic
changes, and new features that may be requested through an enhanced changes, and new features that may be requested through an enhanced
API. API.
TCP features the ability to send "Urgent" data, the delivery of which TCP features the ability to send "Urgent" data, the delivery of which
to the application may or may not be out-of-band. The use of this to the application may or may not be out-of-band. The use of this
feature is not recommended due to security implications and feature is not recommended due to security implications and
implementation differences [20]. MPTCP requires contiguous data to implementation differences [20]. MPTCP requires contiguous data to
support its Data Sequence Mapping over multiple segments, and support its data sequence mapping over multiple segments, and
therefore the Urgent pointer cannot interrupt an existing mapping. therefore the Urgent pointer cannot interrupt an existing mapping.
An MPTCP implementation MAY choose to support sending Urgent data, An MPTCP implementation MAY choose to support sending Urgent data,
and if it does, it SHOULD send the Urgent data on the soonest and if it does, it SHOULD send the Urgent data on the soonest
available unassigned subflow sequence space. Incoming Urgent data available unassigned subflow sequence space. Incoming Urgent data
SHOULD be mapped to connection-level sequence space and delivered to SHOULD be mapped to connection-level sequence space and delivered to
the application analogous to Urgent data in regular TCP. the application analogous to Urgent data in regular TCP.
6.2. Interactions with Management Systems 6.2. Interactions with Management Systems
To enable interactions between TCP and network management systems, To enable interactions between TCP and network management systems,
the TCP [21] and TCP Extended Statistics (ESTATS) [22] MIBs have been the TCP [21] and TCP Extended Statistics (ESTATS) [22] MIBs have been
defined. MPTCP should share the these MIBs for aspects that are defined. MPTCP should share these MIBs for aspects that are designed
designed to be transparent to the application. to be transparent to the application.
It is anticipated that a MPTCP MIB will be defined in the future, It is anticipated that an MPTCP MIB will be defined in the future,
once experience of experimental MPTCP deployments is gathered. This once experience of experimental MPTCP deployments is gathered. This
MIB would provide access to MPTCP-specific properties such as whether MIB would provide access to MPTCP-specific properties such as whether
MPTCP is enabled, and the number and properties of the individual MPTCP is enabled and the number and properties of the individual
paths in use. paths in use.
7. Interactions with Middleboxes 7. Interactions with Middleboxes
As discussed in Section 2.2, it is a goal of MPTCP to be deployable As discussed in Section 2.2, it is a goal of MPTCP to be deployable
today and thus compatible with the majority of middleboxes. This today and thus compatible with the majority of middleboxes. This
section summarises the issues that may arise with NATs, firewalls, section summarizes the issues that may arise with NATs, firewalls,
proxies, intrusion detection systems, and other middleboxes that, if proxies, intrusion detection systems, and other middleboxes that, if
not considered in the protocol design, may hinder its deployment. not considered in the protocol design, may hinder its deployment.
This section is intended primarily as a description of options and This section is intended primarily as a description of options and
considerations only. Protocol-specific solutions to these issues considerations only. Protocol-specific solutions to these issues
will be given in the companion documents. will be given in the companion documents.
Multipath TCP will be deployed in a network that no longer provides Multipath TCP will be deployed in a network that no longer provides
just basic datagram delivery. A myriad of middleboxes are deployed just basic datagram delivery. A myriad of middleboxes are deployed
to optimize various perceived problems with the Internet protocols: to optimize various perceived problems with the Internet protocols:
NATs primarily address IP address space shortage [15], Performance NATs primarily address IP address space shortage [15], Performance
Enhancing Proxies (PEPs) optimize TCP for different link Enhancing Proxies (PEPs) optimize TCP for different link
characteristics [17], firewalls [16] and intrusion detection systems characteristics [17], firewalls [16] and intrusion detection systems
try to block malicious content from reaching a host, and traffic try to block malicious content from reaching a host, and traffic
normalizers [23] ensure a consistent view of the traffic stream to normalizers [23] ensure a consistent view of the traffic stream to
Intrusion Detection Systems (IDS) and hosts. Intrusion Detection Systems (IDS) and hosts.
All these middleboxes optimize current applications at the expense of All these middleboxes optimize current applications at the expense of
future applications. In effect, future applications will often need future applications. In effect, future applications will often need
to behave in a similar fashion to existing ones, in order to increase to behave in a similar fashion to existing ones, in order to increase
the chances of successful deployment. Further, the precise behaviour the chances of successful deployment. Further, the precise behavior
of all these middleboxes is not clearly specified, and implementation of all these middleboxes is not clearly specified, and implementation
errors make matters worse, raising the bar for the deployment of new errors make matters worse, raising the bar for the deployment of new
technologies. technologies.
The following list of middlebox classes documents behaviour that The following list of middlebox classes documents behavior that could
could impact the use of MPTCP. This list is used in [5] to describe impact the use of MPTCP. This list is used in [5] to describe the
the features of the MPTCP protocol that are used to mitigate the features of the MPTCP protocol that are used to mitigate the impact
impact of these middlebox behaviours. of these middlebox behaviors.
o NATs: Network Address Translators decouple the host's local IP o NATs: Network Address Translators decouple the host's local IP
address (and, in the case of NAPTs, port) with that which is seen address (and, in the case of NAPTs, port) with that which is seen
in the wider Internet when the packets are transmitted through a in the wider Internet when the packets are transmitted through a
NAT. This adds complexity, and reduces the chances of success, NAT. This adds complexity, and reduces the chances of success,
when signalling IP addresses. when signaling IP addresses.
o PEPs: Performance Enhancing Proxies, which aim to improve the o PEPs: Performance Enhancing Proxies, which aim to improve the
performance of protocols over low-performance (e.g. high latency performance of protocols over low-performance (e.g., high-latency
or high error rate) links. As such, they may "split" a TCP or high-error-rate) links. As such, they may "split" a TCP
connection and behaviour such as proactive ACKing may occur, and connection and behavior such as proactive ACKing may occur, and
therefore it is no longer guaranteed that one host is therefore it is no longer guaranteed that one host is
communicating directly with another. PEPs, firewalls or other communicating directly with another. PEPs, firewalls, or other
middleboxes may also change the declared receive window size. middleboxes may also change the declared receive window size.
o Traffic Normalizers: These aim to eliminate ambiguities and o Traffic Normalizers: These aim to eliminate ambiguities and
potential attacks at the network level, and amongst other things potential attacks at the network level, and amongst other things,
are unlikely to permit holes in TCP-level sequence space (which are unlikely to permit holes in TCP-level sequence space (which
has impact on MPTCP's retransmission and subflow sequence has an impact on MPTCP's retransmission and subflow sequence
numbering design choices). numbering design choices).
o Firewalls: on top of preventing incoming connections, firewalls o Firewalls: on top of preventing incoming connections, firewalls
may also attempt additional protection such as sequence number may also attempt additional protection such as sequence number
randomization (so a sender cannot reliably know what TCP sequence randomization (so a sender cannot reliably know what TCP sequence
number the receiver will see). number the receiver will see).
o Intrusion Detection Systems: IDSs may look for traffic patterns to o IDSs: Intrusion Detection Systems may look for traffic patterns to
protect a network, and may have false positives with MPTCP and protect a network and may have false positives with MPTCP and drop
drop the connections during normal operation. Future MPTCP-aware the connections during normal operation. Future MPTCP-aware
middleboxes will require the ability to correlate the various middleboxes will require the ability to correlate the various
paths in use. paths in use.
o Content-aware Firewalls: Some middleboxes may actively change data o Content-Aware Firewalls: Some middleboxes may actively change data
in packets, such as re-writing URIs in HTTP traffic. in packets, such as rewriting URIs in HTTP traffic.
In addition, all classes of middleboxes may affect TCP traffic in the In addition, all classes of middleboxes may affect TCP traffic in the
following ways: following ways:
o TCP Options: some middleboxes may drop packets with unknown TCP o TCP Options: some middleboxes may drop packets with unknown TCP
options, or strip those options from the packets. options or strip those options from the packets.
o Segmentation and Coalescing: middleboxes (or even something as o Segmentation and Coalescing: middleboxes (or even something as
close to the end host as TCP Segmentation Offloading (TSO) on a close to the end host as TCP Segmentation Offloading (TSO) on a
Network Interface Card (NIC)) may change the packet boundaries Network Interface Card (NIC)) may change the packet boundaries
from those which the sender intended. It may do this by splitting from those that the sender intended. It may do this by splitting
packets, or coalescing them together. This leads to two major packets or coalescing them together. This leads to two major
impacts: we cannot guarantee where a packet boundary will be, and impacts: where a packet boundary will be cannot be guaranteed and
we cannot say for sure what a middlebox will do with TCP options what a middlebox will do with TCP options in these cases (they may
in these cases (they may be repeated, dropped, or sent only once). be repeated, dropped, or sent only once) cannot be said for sure.
8. Contributors 8. Contributors
The authors would like to acknowledge the contributions of Andrew The authors would like to acknowledge the contributions of Andrew
McDonald and Bryan Ford to this document. McDonald and Bryan Ford to this document.
The authors would also like to thank the following people for The authors would also like to thank the following people for
detailed reviews: Olivier Bonaventure, Gorry Fairhurst, Iljitsch van detailed reviews: Olivier Bonaventure, Gorry Fairhurst, Iljitsch van
Beijnum, Philip Eardley, Michael Scharf, Lars Eggert, Cullen Beijnum, Philip Eardley, Michael Scharf, Lars Eggert, Cullen
Jennings, Joel Halpern, Juergen Quittek, Alexey Melnikov, David Jennings, Joel Halpern, Juergen Quittek, Alexey Melnikov, David
Harrington, Jari Arkko and Stewart Bryant. Harrington, Jari Arkko, and Stewart Bryant.
9. Acknowledgements 9. Acknowledgements
Alan Ford, Costin Raiciu, Mark Handley, and Sebastien Barre are Alan Ford, Costin Raiciu, Mark Handley, and Sebastien Barre are
supported by Trilogy (http://www.trilogy-project.org), a research supported by Trilogy (http://www.trilogy-project.org), a research
project (ICT-216372) partially funded by the European Community under project (ICT-216372) partially funded by the European Community under
its Seventh Framework Program. The views expressed here are those of its Seventh Framework Program. The views expressed here are those of
the author(s) only. The European Commission is not liable for any the author(s) only. The European Commission is not liable for any
use that may be made of the information in this document. use that may be made of the information in this document.
10. IANA Considerations 10. Security Considerations
None.
11. Security Considerations
This informational document provides an architectural overview for This informational document provides an architectural overview for
Multipath TCP and so does not, in itself, raise any security issues. Multipath TCP and so does not, in itself, raise any security issues.
A separate threat analysis [12] lists threats that can exist with a A separate threat analysis [12] lists threats that can exist with a
Multipath TCP. However, a protocol based on the architecture in this Multipath TCP. However, a protocol based on the architecture in this
document will have a number of security requirements. The high level document will have a number of security requirements. The high-level
goals for such a protocol are identified in Section 2.3, whilst goals for such a protocol are identified in Section 2.3, whilst
Section 5.8 provides more detailed discussion of security Section 5.8 provides more detailed discussion of security
requirements and design decisions which are applied in the MPTCP requirements and design decisions which are applied in the MPTCP
protocol design [5]. protocol design [5].
12. References 11. References
12.1. Normative References 11.1. Normative References
[1] Postel, J., "Transmission Control Protocol", STD 7, RFC 793, [1] Postel, J., "Transmission Control Protocol", STD 7, RFC 793,
September 1981. September 1981.
[2] Bradner, S., "Key words for use in RFCs to Indicate Requirement [2] Bradner, S., "Key words for use in RFCs to Indicate Requirement
Levels", BCP 14, RFC 2119, March 1997. Levels", BCP 14, RFC 2119, March 1997.
12.2. Informative References 11.2. Informative References
[3] Wischik, D., Handley, M., and M. Bagnulo Braun, "The Resource [3] Wischik, D., Handley, M., and M. Bagnulo Braun, "The Resource
Pooling Principle", ACM SIGCOMM CCR vol. 38 num. 5, pp. 47-52, Pooling Principle", ACM SIGCOMM CCR vol. 38 num. 5, pp. 47-52,
October 2008, October 2008,
<http://ccr.sigcomm.org/online/files/p47-handleyA4.pdf>. <http://ccr.sigcomm.org/online/files/p47-handleyA4.pdf>.
[4] Hopps, C., "Analysis of an Equal-Cost Multi-Path Algorithm", [4] Hopps, C., "Analysis of an Equal-Cost Multi-Path Algorithm",
RFC 2992, November 2000. RFC 2992, November 2000.
[5] Ford, A., Raiciu, C., and M. Handley, "TCP Extensions for [5] Ford, A., Raiciu, C., Handley, M., and O. Bonaventure, "TCP
Multipath Operation with Multiple Addresses", Extensions for Multipath Operation with Multiple Addresses",
draft-ietf-mptcp-multiaddressed-02 (work in progress), Work in Progress, March 2011.
October 2010.
[6] Stewart, R., "Stream Control Transmission Protocol", RFC 4960, [6] Stewart, R., "Stream Control Transmission Protocol", RFC 4960,
September 2007. September 2007.
[7] Raiciu, C., Handley, M., and D. Wischik, "Coupled Congestion [7] Raiciu, C., Handley, M., and D. Wischik, "Coupled Congestion
Control for Multipath Transport Protocols", Control for Multipath Transport Protocols", Work in Progress,
draft-ietf-mptcp-congestion-01 (work in progress), March 2011.
January 2011.
[8] Scharf, M. and A. Ford, "MPTCP Application Interface [8] Scharf, M. and A. Ford, "MPTCP Application Interface
Considerations", draft-ietf-mptcp-api-00 (work in progress), Considerations", Work in Progress, March 2011.
November 2010.
[9] Carpenter, B. and S. Brim, "Middleboxes: Taxonomy and Issues", [9] Carpenter, B. and S. Brim, "Middleboxes: Taxonomy and Issues",
RFC 3234, February 2002. RFC 3234, February 2002.
[10] Carpenter, B., "Internet Transparency", RFC 2775, [10] Carpenter, B., "Internet Transparency", RFC 2775,
February 2000. February 2000.
[11] Mathis, M., Mahdavi, J., Floyd, S., and A. Romanow, "TCP [11] Mathis, M., Mahdavi, J., Floyd, S., and A. Romanow, "TCP
Selective Acknowledgment Options", RFC 2018, October 1996. Selective Acknowledgment Options", RFC 2018, October 1996.
[12] Bagnulo, M., "Threat Analysis for TCP Extensions for Multi-path [12] Bagnulo, M., "Threat Analysis for TCP Extensions for Multipath
Operation with Multiple Addresses", draft-ietf-mptcp-threat-07 Operation with Multiple Addresses", RFC 6181, March 2011.
(work in progress), January 2011.
[13] Becke, M., Dreibholz, T., Iyengar, J., Natarajan, P., and M. [13] Becke, M., Dreibholz, T., Iyengar, J., Natarajan, P., and M.
Tuexen, "Load Sharing for the Stream Control Transmission Tuexen, "Load Sharing for the Stream Control Transmission
Protocol (SCTP)", draft-tuexen-tsvwg-sctp-multipath-01 (work in Protocol (SCTP)", Work in Progress, December 2010.
progress), December 2010.
[14] Ford, B. and J. Iyengar, "Breaking Up the Transport Logjam", [14] Ford, B. and J. Iyengar, "Breaking Up the Transport Logjam",
ACM HotNets, October 2008. ACM HotNets, October 2008.
[15] Srisuresh, P. and K. Egevang, "Traditional IP Network Address [15] Srisuresh, P. and K. Egevang, "Traditional IP Network Address
Translator (Traditional NAT)", RFC 3022, January 2001. Translator (Traditional NAT)", RFC 3022, January 2001.
[16] Freed, N., "Behavior of and Requirements for Internet [16] Freed, N., "Behavior of and Requirements for Internet
Firewalls", RFC 2979, October 2000. Firewalls", RFC 2979, October 2000.
skipping to change at page 28, line 5 skipping to change at page 28, line 5
Transmission Control Protocol (TCP)", RFC 4022, March 2005. Transmission Control Protocol (TCP)", RFC 4022, March 2005.
[22] Mathis, M., Heffner, J., and R. Raghunarayan, "TCP Extended [22] Mathis, M., Heffner, J., and R. Raghunarayan, "TCP Extended
Statistics MIB", RFC 4898, May 2007. Statistics MIB", RFC 4898, May 2007.
[23] Handley, M., Paxson, V., and C. Kreibich, "Network Intrusion [23] Handley, M., Paxson, V., and C. Kreibich, "Network Intrusion
Detection: Evasion, Traffic Normalization, and End-to-End Detection: Evasion, Traffic Normalization, and End-to-End
Protocol Semantics", Usenix Security 2001, 2001, <http:// Protocol Semantics", Usenix Security 2001, 2001, <http://
www.usenix.org/events/sec01/full_papers/handley/handley.pdf>. www.usenix.org/events/sec01/full_papers/handley/handley.pdf>.
Appendix A. Changelog
(For removal by the RFC Editor)
A.1. Changes since draft-ietf-mptcp-architecture-04
o Responded to IETF Last Call and IESG review comments.
A.2. Changes since draft-ietf-mptcp-architecture-03
o Responded to AD review comments.
A.3. Changes since draft-ietf-mptcp-architecture-02
o Responded to WG last call review comments. Included editorial
fixes, adding Section 2.4, and improving Section 5.4 and
Section 7.
A.4. Changes since draft-ietf-mptcp-architecture-01
o Responded to review comments.
o Added security sections.
A.5. Changes since draft-ietf-mptcp-architecture-00
o Added middlebox compatibility discussion (Section 7).
o Clarified path identification (TCP 4-tuple) in Section 5.5.
o Added brief scenario and diagram to Section 1.3.
Authors' Addresses Authors' Addresses
Alan Ford Alan Ford
Roke Manor Research Roke Manor Research
Old Salisbury Lane Old Salisbury Lane
Romsey, Hampshire SO51 0ZN Romsey, Hampshire SO51 0ZN
UK UK
Phone: +44 1794 833 465 Phone: +44 1794 833 465
Email: alan.ford@roke.co.uk EMail: alan.ford@roke.co.uk
Costin Raiciu Costin Raiciu
University College London University College London
Gower Street Gower Street
London WC1E 6BT London WC1E 6BT
UK UK
EMail: c.raiciu@cs.ucl.ac.uk
Email: c.raiciu@cs.ucl.ac.uk
Mark Handley Mark Handley
University College London University College London
Gower Street Gower Street
London WC1E 6BT London WC1E 6BT
UK UK
EMail: m.handley@cs.ucl.ac.uk
Email: m.handley@cs.ucl.ac.uk
Sebastien Barre Sebastien Barre
Universite catholique de Louvain Universite catholique de Louvain
Pl. Ste Barbe, 2 Pl. Ste Barbe, 2
Louvain-la-Neuve 1348 Louvain-la-Neuve 1348
Belgium Belgium
Phone: +32 10 47 91 03 Phone: +32 10 47 91 03
Email: sebastien.barre@uclouvain.be EMail: sebastien.barre@uclouvain.be
Janardhan Iyengar Janardhan Iyengar
Franklin and Marshall College Franklin and Marshall College
Mathematics and Computer Science Mathematics and Computer Science
PO Box 3003 PO Box 3003
Lancaster, PA 17604-3003 Lancaster, PA 17604-3003
USA USA
Phone: 717-358-4774 Phone: 717-358-4774
Email: jiyengar@fandm.edu EMail: jiyengar@fandm.edu
 End of changes. 134 change blocks. 
341 lines changed or deleted 285 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/