A Framework of Multipath Transport System Based on Application-Level Relay (MPTS-AR)
draft-leiwm-tsvwg-mpts-ar-00
The information below is for an old version of the document.
| Document | Type |
This is an older version of an Internet-Draft whose latest revision state is "Expired".
|
|
|---|---|---|---|
| Authors | Lei Weimin , Wei Zhang , Shaowei Liu | ||
| Last updated | 2013-07-29 | ||
| RFC stream | (None) | ||
| Formats | |||
| Stream | Stream state | (No stream defined) | |
| Consensus boilerplate | Unknown | ||
| RFC Editor Note | (None) | ||
| IESG | IESG state | I-D Exists | |
| Telechat date | (None) | ||
| Responsible AD | (None) | ||
| Send notices to | (None) |
draft-leiwm-tsvwg-mpts-ar-00
Network Working Group W. Lei
Internet-Draft W. Zhang
Intended status: Experimental S. Liu
Expires: January 27, 2014 Northeastern University
July 28, 2013
A Framework of Multipath Transport System Based on
Application-Level Relay (MPTS-AR)
draft-leiwm-tsvwg-mpts-ar-00.txt
Abstract
Multipath transport is an important way to improve the efficiency of
data delivery. This document defines a multipath transport system
framework in which application-level relays are deployed to provide
the conditions to enable multiple paths between source and
destination. In the proposed framework, endpoints are allowed to use
multiple paths, including the default IP path and relay paths, to
transport data in a single session. A relay path may via one or more
relays which provide application-level relay services for endpoints.
This framework defines three kinds of logical entities including user
agent, relay server and relay controller. Relay server provides relay
service for user agents based on a local path-table. Relay controller
manages relay servers and relay paths. User agent maintains multiple
end-to-end paths which include a default path and multiple relay
paths. The framework also defines a relay service control protocol
named OpenPath protocol in control plane to manage relay paths, and a
profile of multipath transport protocol suite in data plane to
facilitate multipath data transport. The multipath transport system
framework can support various applications including applications
requiring timely delivery of real-time data such as streaming media,
and applications requiring ordered reliable delivery of stream of
data such as file transfer.
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
Task Force (IETF). Note that other groups may also distribute
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 and may be updated, replaced, or obsoleted by other documents
at any time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
Leiwm, et al. Expires January 27, 2014 [Page 1]
Internet-Draft A Framework of Multipath Transport System July 2013
This Internet-Draft will expire on January 27, 2014.
Copyright Notice
Copyright (c) 2013 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents (http://trustee.ietf.org/
license-info) in effect on the date of publication of this document.
Please review these documents carefully, as they describe your rights
and restrictions with respect to this document. Code Components
extracted from this document must include Simplified BSD License text
as described in Section 4.e of the Trust Legal Provisions and are
provided without warranty as described in the Simplified BSD License.
Table of Contents
1. Introduction ................................................ 3
2. Terminology ................................................. 5
3. Definitions ................................................. 5
4. Overview .................................................... 7
4.1 Deployment and organization of relay controller and
relay server ........................................... 8
4.2 Relay path service provided by relay controller ........ 9
4.3 End-to-end transmission paths managed by user agent .... 10
4.4 Relay service control protocol ......................... 11
4.5 Multipath transport protocol ........................... 12
5. Use Scenarios ............................................... 13
5.1 Usage Scenario in SIP system ........................... 15
6. User Agent Behavior ......................................... 17
6.1 Multipath Session Management ........................... 17
6.2 Path Management ........................................ 18
6.3 Flow Partitioning and Scheduling ....................... 19
6.4 Subflow Packaging ...................................... 20
6.5 Subflow Recombination .................................. 20
6.6 Subflow Reporting ...................................... 20
7. Relay Server Behavior ....................................... 21
7.1 Connection Management and Registrations ................ 21
7.2 Path-Table Management .................................. 22
7.3 Path Validity Management ............................... 24
7.4 Relay Service Management ............................... 25
7.5 MPTP Packet Processing ................................. 25
8. Relay Controller Behavior ................................... 25
8.1 Relay Server Management ................................ 25
8.2 Relay Path Allocation .................................. 26
9. OpenPath Protocol ........................................... 26
Leiwm, et al. Expires January 27, 2014 [Page 2]
Internet-Draft A Framework of Multipath Transport System July 2013
9.1 Protocol Overview ...................................... 26
9.1.1 Relay-to-Controller ............................. 27
9.1.2 Controller-to-Relay ............................. 28
9.1.3 User Agent-to-Controller ........................ 29
9.1.4 Symmetric ....................................... 29
9.2 Common Structures ...................................... 29
9.2.1 OpenPath Common Header .......................... 29
9.2.2 Common Body of OpenPath Failure Responses ....... 31
9.2.3 Transport Address Structure ..................... 32
9.3 Message Format of OpenPath Request and Success Response. 32
9.3.1 HELLO ........................................... 32
9.3.2 START/STOP/BYE .................................. 33
9.3.3 ECHO ............................................ 33
9.3.4 NOTIFY/DELETE_PATH .............................. 33
9.3.5 ADD_PATH/ UPDATE_PATH ........................... 34
9.3.6 ALLOCATE_PATH/RELEASE_PATH ...................... 35
9.3.7 FEATURES ........................................ 35
9.3.8 STATISTICS ...................................... 35
10. MPTP Profile ................................................ 35
10.1 Overview ............................................... 35
10.2 MPTP Fixed Header Fields ............................... 36
10.2.1 Fixed Header Fields of MPTP Data Packet ........ 36
10.2.2 Fixed Header Fields of MPTP Control Packet ..... 38
11. SDP Considerations .......................................... 39
11.1 Signaling MPTP Capability in SDP ....................... 39
11.2 Relay Path Advertisement in SDP ........................ 40
12. IANA Considerations ......................................... 41
12.1 SDP Attributes ......................................... 41
13. Security Considerations ..................................... 42
14. References .................................................. 42
14.1 Normative References ................................... 42
14.2 Informative References ................................. 42
1. Introduction
For end-to-end multimedia session, multipath transport has more
benefits than single path transmission. Multiple disjoint (or
partially disjoint) paths could provide greater transmission
bandwidth, and redundancy among multiple paths increases transmission
reliability by protecting multiple paths from failure of one, so
multipath transport can promote guarantee ability of transmission
service of quality. Moreover, from the perspective of whole network
transmission efficacy, multipath transport can achieve load balance
and increase the efficiency of the network resource usage.
In order to achieve multipath transport, the following two problems
need to be considered: 1) how to build multiple paths between
communication endpoints, and 2) given multipath paths, how to
Leiwm, et al. Expires January 27, 2014 [Page 3]
Internet-Draft A Framework of Multipath Transport System July 2013
implement multipath transport between communication endpoints.
Current IP protocol can only build a single transport path between
endpoints and it is nearly impossible to update protocol stack of
existing network devices to support multipath routing. Now the main
scenario of multipath transport is based on multi-homed host, which
requires at least one of communicating endpoints is multi-homed. The
multi-homed host multipath scenario relies on the end host equipped
with several access networks which is hard to be satisfied.
Establishing a multipath transport scenario based on application-
level relay is one of application-level multipath routing methods.
It can establish multiple disjoint (partially disjoint) transmission
paths which pass one or more relay servers between end hosts. This
scheme of multipath transport is a kind of overlay network
technologies actually. It is compatible with existing protocol stack,
and gets rid of the restrictions of physical network conditions.
Therefore, it is relatively easier to establish multipath transport
scenario.
This document defines the framework of multipath transport system
based on application-level relay. A large amount of application-level
relays are deployed to provide relay service to the communicating
endpoints. The upper application programs are provided with
opportunities to autonomously select one or multiple paths, including
the default IP path and relay paths, to transport data in a session.
This framework defines three kinds of logical entities including user
agent, relay server and relay controller, and describes their
function blocks and behaviors. The framework also defines a relay
service control protocol named OpenPath protocol in control plane to
manage relay paths, and a profile of multipath transport protocol
suite in data plane to facilitate multipath data transport.
Relay controller and relay servers constitute a relay service system,
which provides relay service to user agents. Relay controller is
responsible for managing relay servers, allocating relay paths, QoS
condition evaluation and so on. It also provides a unified access
interface of relay path service to user agent or out-of-band
signaling entity. Main function of relay server is to complete the
application-level data forwarding, by receiving media stream from
the source address and port, and then forwarding to the address and
port of next hop. A relay path may pass one or more relay servers.
OpenPath mainly includes two kinds of message: the first type is
the message between relay controller and relay server which is used
to manage the relay server. Any application software or special
server, which implements data relay services and supports OpenPath
protocol, can dynamically register to a relay controller and provide
Leiwm, et al. Expires January 27, 2014 [Page 4]
Internet-Draft A Framework of Multipath Transport System July 2013
relay service for user agents in the region of this relay controller.
The second type is the message between relay controller and user
agent or out-of-band signaling entity which is used to manage relay
path including relay path inquiry, establishment, teardown, etc.
In order to support various applications, the multipath transport
system framework based on application-level relay needs to work
with a multipath transport protocol (MPTP) suite which consists of
multiple application-specific MPTPs. Each application-specific MPTP
is aimed to meet the transmission requirements of a specific
category of upper applications. All application-specific MPTPs MUST
follow the common rules defined by MPTP profile which provides
application-level multipath routing mechanism. This document gives
the definition of MPTP profile. Application-specific MPTPs are
defined in companion documents.
Main principles for designing this framework are:
1) Establishment of multipath transport scenarios depends on the
relay service system which does not care end-to-end transmission
characteristics. Therefore, a variety of upper applications can use
multipath transport services provided by the relay service system.
2) Relay service management and service access interfaces are
standardized so that any organizations and individuals can carry out
specialized relay transmission services.
3) The MPTP profile defines common rules of multipath transport based
on application-level relay for various upper applications. To support
a specific category of upper applications, a corresponding
application-specific MPTP should be defined in an additional document.
2. Terminology
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in RFC2119 [1].
3. Definitions
1) Subflow: flow of data packets along a specific path, i.e., a
subset of the packets belonging to a data flow. The combination of
all subflows forms the complete original data flow. Typically, a
subflow should map to a unique path.
2) Session: an association between two participants transmitting
data. An endpoint may be involved in multiple sessions at the same
time. For example, in a multimedia session, each medium is typically
Leiwm, et al. Expires January 27, 2014 [Page 5]
Internet-Draft A Framework of Multipath Transport System July 2013
carried in a separate real-time transport protocol (RTP) [3] session
which is a type of 'session' defined here. Another typical example of
session is to transfer one or multiple files between two endpoints.
This framework allows the variations defined here for different
applications.
3) Multipath Session: a special type of session in which the
original data flow is split into multiple subflows and each subflow
is mapped to a unique path.
4) Path: sequence of logical links between a source and a
destination. We define it by a set of addresses. All available paths
for a session include a default path and one or more relay paths.
5) Default path: a path between a sender and a receiver, which is
same as the path negotiated and established by a normal session. In
Session Initiation Protocol (SIP) [4] and Session Description
Protocol (SDP) [5] case, the default path is determined by the c= and
m= lines in SDP during session setup. If either the source or the
destination is not behind a symmetric NAT, the default path may be
the direct network path between the source and the destination.
Otherwise, it may traverse a third-party node, such as a TURN server
or a media server which is responsible for relaying packets.
6) Relay path: a path via one or multiple relay servers between a
source and a destination. A relay path is defined by a sequence of
(S, R1, ..., Rm, D), where Ri denotes the address of the i-th relay
server and m denotes the number of relay servers in this path.
7) Candidate path: a path that is either a default path or a relay
path.
8) Active path: a path that carries media data.
9) User Agent: a logical entity that can act as either a sender or
a receiver. A sender is responsible for managing all candidate paths
and multipath session, partitioning and scheduling subflows, and
packaging MPTP packets. A receiver is responsible for recombinating
subflow packets and sending back QoS feedback for each subflow. The
behavior of a user agent is further defined in Section 6.
10) Relay server: an intermediary entity that primarily plays the
role of forwarding subflow packets according to a Path-Table. Relay
server receives subflow packets from its upstream entity of the
subflow that the received packets belong to, obtains the next-hop
transport address based on the matching results in the Path-Table,
and forwards the packets to the corresponding next-hop transport
address. The behavior of a relay server is further defined in Section
Leiwm, et al. Expires January 27, 2014 [Page 6]
Internet-Draft A Framework of Multipath Transport System July 2013
7.
11) Path-Table: a table consisting of a set of path entries, which
may be added, updated or deleted by a remote relay controller. Each
relay server maintains its own path-table.
12) Relay Controller: a logical entity that manages all relay
servers in its region and allocates relay paths for each session. The
behavior of a relay controller is further defined in Section 8.
4. Overview
In the multipath transport system based on application-level relay,
relay server provides application-level data forwarding service,
which can used to establish multiple relay paths between
communication endpoints to achieve multipath transport. With
appropriate multipath transport control mechanisms and protocols, not
only can it achieve higher quality of end-to-end transmission
service, but it balance network load and enhance efficiency of the
network resource usage.
Figure 1 illustrates the structure of the proposed multipath
transport system framework and the relationship between user agent,
relay server and relay controller. Relay controller and relay server
constitute the relay service system which provides relay service to
user agent. User agent maintains multiple end-to-end relay paths,
and dispatches data flow along multipath paths following MPTP
protocol suite.
Leiwm, et al. Expires January 27, 2014 [Page 7]
Internet-Draft A Framework of Multipath Transport System July 2013
(1)Out-of-band Signaling +-----------------+ (1)
+--------------------->| Out-of-band |<------------------+
| | Signaling Server| |
| +-----------------+ |
| ^ |
| |(1)OpenPath |
| V |
| or (2) OpenPath +------------------+ |
| +----------------->| Relay Controller | |
| | +------------------+ |
| | ^ |
| | | |
V V | V
+------------+ |OpenPath +------------+
| User Agent | | | User Agent |
| (Sender) | | | (Receiver) |
+------------+ V +------------+
| ................................. ^
| . +-----------------+ . |
| . | Relay server | . |
| . +-----------------+ | . |
| . | Relay server |--+ . |
| . +-----------------+ | . |
+------------>. | Relay server |--+ .------------+
MPTP . | | . MPTP
. +-----------------+ .
Figure 1. The structure of the multipath transport system framework
based on application-level relay
4.1 Deployment and organization of relay controller and relay server
The relay controller is the central component of the relay service
system. The main functions of relay controller are to manage relay
servers, to evaluate the QoS condition of relay paths, to provide
relay path service to user agents (or through out-of-band signaling
server provide service to user agent indirectly). Relay controller is
usually be deployed in the network by the application service
provider. When there are a large number of user agents, multiple
controllers can be deployed and each of them provides management
service for part of user agents.
Relay servers provide data relay service to the communicating agents.
Considering the efficiency of the forwarding service, multipath
transport system framework based on application-level relay adopts
UDP as the underlying transport protocol. Relay servers may be
deployed in a variety of ways. Combined with application
requirements, application service providers can deploy a large number
Leiwm, et al. Expires January 27, 2014 [Page 8]
Internet-Draft A Framework of Multipath Transport System July 2013
of proprietary servers with high network bandwidth and computing
performance in their system. The participating nodes that access the
same application may also self-organize to form a dynamic overlay
network among which the nodes with higher performance can provide
relay service to others. If relay server has a stronger forwarding
ability and bigger scale, the relay transmission would have better
path diversity and better QoS. The organization of relay servers is a
kind of typical overlay network technology. Overlay network model of
relay servers is closely related to the generation and selection of
relay paths so that this document does not recommend any specific
overlay network model, and only rules that all relay servers need to
register to a relay controller and provide data relay service to user
agents.
4.2 Relay path service provided by relay controller
All relay servers register to a relay controller in their region.
Relay controller is responsible for managing the relay servers and
providing direct or indirect relay path services to the user agent by
OpenPath protocol including establishing inquiring and removing of
relay paths.
Considering the execution efficiency, the relay path is designed to
be unidirectional. A bidirectional data flow, such as in a
conversational use-case, is regarded as two independent
unidirectional flows in opposite directions. Relay paths
are assigned for each unidirectional flow.
User agent sender, who is responsible for sending out data flow, and
relay servers need to know the relay path information so they can
correctly forward subflow data along a particular relay path. An
alternative approach, similar to source routing, is that the user
agent sender can store the entire route in MPTP packet headers. Each
intermediate node will simply examine the headers of a received
packet and forward it to the next node as indicated in the source
route. The advantage of the method is to simplify the implementation
of relay servers. They need not store any path information and can
perform routing of MPTP packets only based on MPTP packet headers.
But this method introduces traffic overhead considerably, especially
when the payload traffic is relatively small.
In practice, the user agent sender and relay servers need not to know
the complete information of the associated path. They just need to
know the next-hop transport address for each path associated with
them. A pair of transport address comprises one network address plus
one port. When receiving OpenPath path allocation request messages
from user agents directly or indirectly (for example via signaling
server such as SIP or RTSP), relay controller generates a set of
Leiwm, et al. Expires January 27, 2014 [Page 9]
Internet-Draft A Framework of Multipath Transport System July 2013
disjoint relay paths based on information carried by the OpenPath
path allocation request message including the media types, quality of
experience (QoE) expectations or requirements, the number of relay
paths expected, and so on. Relay controller associates a unique path
identifier to each relay path. On the one hand, relay controller
instructs the corresponding relay servers to assign the appropriate
resources for incoming data forwarding. On the other hand, relay
controller sends OpenPath response message back to the service
requester. Usually, the response message includes the path
identifier, the relay service address of first-hop relay server and
instant QoS of each path. This document neither defines a uniform
relay path generation algorithm, nor recommends QoS evaluation method
of relay path. These methods can be designed by the application
service provider. In addition, considering the operability of QoS
evaluation of relay path and the efficiency of relaying processing,
all relay paths are unidirectional.
Relay controller need to maintain the correspondence between the user
agent and the allocated relay paths. When receiving OpenPath path
removal request messages, relay controller instructs the associated
relay servers to release resources and generates a response message
back to the requester.
4.3 End-to-end transmission paths managed by user agent
End-to-end transmission path between user agents includes two types:
1) default path (DP), which does not via relay server; 2) relay path
(RP), which via one or more relay servers.
In order to establish multiple end-to-end transmission paths, user
agent sender, needs to collect candidate paths before transmitting
data. As shown in figure 1, this document provides two alternative
ways to be used for collecting candidate paths by the user agent
sender. In the first way, the user agent sender obtains candidate
paths from the out-of-band signaling used for establishing an
association between the user agent sender and user agent receiver.
More specifically, user agents use a kind of out-of-band signaling
(e.g., SIP, Real-Time Streaming Protocol (RTSP) [6]) to negotiate
and establish a session with the remote peer before transmitting
data. The signaling server, which out-of-band signaling passes
through during session setup, is extended to support the access
interfaces of relay path service provided by OpenPath protocol. The
signaling server requests the relay controller to allocate relay
paths for the session to be established and inserts the information
of the allocated relay paths into corresponding signaling messages
to inform the participating user agents. In the second way, the user
gent sender obtains candidate paths through direct interaction with
the relay controller using OpenPath Protocol. The advantage of the
Leiwm, et al. Expires January 27, 2014 [Page 10]
Internet-Draft A Framework of Multipath Transport System July 2013
first way is to avoid a large number of connections of the relay
controller with user agents, and improve the security of the relay
controller through limiting communication only with the trusted
signaling server.
After collecting multiple end-to-end transmission paths, user agent
sender needs to evaluate the quality of the part of a relay path
from the user agent sender itself to the first-hop relay server
(the first relay server on a relay path) for each relay path.
Combined with path QoS information carried by OpenPath path
allocation response message, user agent sender then calculates
end-to-end transmission qualities of all relay paths, which are used
as the basis of subsequent path selection and load distribution.
During the lifecycle of multipath transport, user agent sender
usually needs to dynamically evaluate QoS condition of each path
using MPTP, and achieves a better transmission service quality
through optimization of dynamic load distribution.
User agents should establish a multipath session before using
multiple paths to transport data flow. It can be set up in many
possible ways e.g., during session establishment, or at anytime
during the session. To reduce session startup time, the user agent
sender can start transporting data via the default path and then
perform path connectivity checks for relay paths. If there are one
or multiple available relay paths, the use agent sender updates the
single-path session to a multipath session.
4.4 Relay service control protocol
Different from multipath transport scenario based on multi-homed
host, in the multipath transport scenario based on application-level
relay, the generation and normal operations of relay paths depend on
the relay service system which includes relay controller and relay
server.
Relay controller provides control functions of relay service, and
relay sever provides executive functions of relay service. Relay
server needs to report the status to the relay controller, and relay
controller needs to control the behaviors of relay servers. In
addition, in order to provide relay path service to user agents or
out-of-band signaling server, relay controller needs to provide relay
service access interfaces. Therefore, this document defines a relay
service control protocol named OpenPath, which provides interfaces
among relay controller, relay server and user agents or out-of-band
signaling server.
The definition of OpenPath protocol will help standardize the
Leiwm, et al. Expires January 27, 2014 [Page 11]
Internet-Draft A Framework of Multipath Transport System July 2013
application-level relay service. As long as OpenPath protocol and
MPTP are followed, third-party relay servers implemented and deployed
by any organization and individual can interact with relay controller
and provide relay service to user agents. This document defines
OpenPath protocol in detail.
4.5 Multipath transport protocol
Multipath transport protocol (MPTP) is an application-layer
transport protocol. In order to be suitable for various
applications, MPTP is designed to be a protocol suite which consists
of one MPTP profile and multiple application-specific MPTPs. The
MPTP profile is aimed to provide application-level multipath routing
mechanism and each application-specific MPTP is aimed to meet the
transmission requirements of a specific category of upper
applications. All application-specific MPTPs MUST follow MPTP
profile defined in this document. Relay servers need to support MPTP
profile, and user agents need to support MPTP profile and one or
multiple application-specific MPTPs.
MPTP is a UDP-based application-layer transport protocol. The
protocol stack architecture of MPTP is shown in figure 2.
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Applications |
|(VoIP, video conference, streaming, file transfer,...) |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| MPTP |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| UDP |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| IP |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Figure 2. The protocol stack architecture of MPTP
In network protocol stack, the transport layer provides end-to-end
communication services for applications. Reliable transport
protocols, such as TCP and SCTP, are not particularly suitable for
real-time applications such as streaming media and real-time
multiplayer games. And they are complex and can be a problem for
relay servers which provide relay service for huge numbers of user
agents. In contrast, UDP only provides checksums for data integrity,
and multiplexing for different applications. It delegates other
functions to the above application programs. Therefore it typically
gives higher throughput and shorter latency. Based on these
considerations, MPTP uses UDP as underlying transport protocol. Each
UDP datagram carries one MPTP packet. On the one hand, this design
Leiwm, et al. Expires January 27, 2014 [Page 12]
Internet-Draft A Framework of Multipath Transport System July 2013
decision simplifies the behaviors and enhances service capabilities
of relay servers. In other words, relay server works in a stateless
manner and provides the UDP-based relay service. On the other hand,
all upper applications, that need either timely delivery or reliable
delivery, can use the transport service provided by MPTP. The
delivery requirements of upper application need to be meet by MPTP
profile and application-specific MPTPs.
In addition to carrying payload data passed from upper application
programs through multiple paths, MPTP also need to provide path
control functions including keeping path alive and monitoring the
quality of data delivery on each path.
5. Usage Scenarios
The multipath transport system framework based on application-level
relay can provide many application scenarios including point-to-
point, many-to-one and one-to-many communications. These
applications could be delay sensitive or reliability sensitive, such
as real-time communication, parallel streaming media, file transfer
or file sharing.
Figure 3 illustrates a point-to-point multipath session. There are
three paths between source and destination, including the default
path, one relay path via one relay server and another relay path via
two relay servers. User agent sender can choose a data partitioning
method according to its particular requirements. Then, each flow is
assigned to a path. The user agent receiver reassembles the received
flows using a resequencing buffer to retrieve the reconstructed flow
which is delivered to the above application.
Relay path(A, Relay-1, B)
+-----------+
+-------------------->| Relay-1 |-----------------------+
| +-----------+ |
| V
+--------+ Default path(A,B) +--------+
| User A |<----------------------------------------------->| User B |
+--------+ +--------+
| ^
| Relay path(A, Relay-2, Relay-3, B) |
| +-----------+ +-----------+ |
+---->| Relay-2 |-------------------->| Relay-3 |-----+
+-----------+ +-----------+
Figure 3. A point-to-point multipath session
A wide range of applications require data transmission from
Leiwm, et al. Expires January 27, 2014 [Page 13]
Internet-Draft A Framework of Multipath Transport System July 2013
geographically distributed sources to one destination. For instance,
large volume data of high definition movies are stored at multiple
geographically distributed locations. The audiences need to retrieve
and integrate data from several locations. This usage scenario can be
completed by a many-to-one multipath session, which is depicted in
Figure 4. In the figure, the user agent receiver streams different
portions of a video from three servers concurrently. The path between
the server and the user agent receiver may go through one or more
relay servers.
+---+
| A |----------------------------------------------------+
+---+ |
|
V
+---+ +---------+ +---+
| B |------------------| Relay |-------------------->| D |
+---+ +---------+ +---+
^
|
+---+ +---------+ |
| C |------------------| Relay |-----------------------+
+---+ +---------+
Figure 4. A many-to-one multipath session
Many video applications are typically characterized by a wide range
of connection qualities and receiving devices which are with
different capabilities ranging from cell phones with small screens
and restricted processing power to high-end PC with high-definition
display. These systems are usually adopting layered coding. Layered
coding enables the encoding of a high-quality video bit stream
containing one or more subset bit streams that can themselves be
decoded independently. This usage scenario can be completed by a
one-to-many multipath session, which is depicted in figure 5. A video
source is encoded into multiple streams, each of which is transported
by a source tree for video multicast.
Leiwm, et al. Expires January 27, 2014 [Page 14]
Internet-Draft A Framework of Multipath Transport System July 2013
+---+
+------------------------->| A |
| +--->+---+
| |
+---------+ |
+------------------->|RTP Relay|----------------|------+
| +---------+ | |
| | | |
| | | V
+---+ +------------------+ | +---+
| S | | | | B |
+---+ +------------------|--+ +---+
| | | ^
| | | |
| +---------+ | |
+------------------->|RTP Relay|-------------|---------+
+---------+ |
| |
| +------>+---+
+------------------------->| C |
+---+
Figure 5. A one-to-many multipath session
5.1 Usage Scenario in SIP system
Figure 6 depicts a kind of usage scenario in which SIP is used as a
separate signaling to advertise relay paths information.
Leiwm, et al. Expires January 27, 2014 [Page 15]
Internet-Draft A Framework of Multipath Transport System July 2013
(1)INVITE +-----------------+ (3)INVITE
+---------------------->| |------------------------+
| +-------------------| SIP Server |<-------------------+ |
| | (6)200OK +-----------------+ (4)200OK | |
| | ^ | |
| | |(2)(5) | |
| | V | |
| | +-----------------+ | |
| | | Relay Controller| | |
| | +-----------------+ | |
| | ^ ^ ^ | |
| | (2)(5) | |(2)(5) | (2)(5) | |
| V +-------+ V +-------+ | V
+--------+ P-1/P-3 | +------------+ | P-1/P-3 +--------+
| User A |<---------|------>| Relay-1 |<-------|----------->| User B |
+--------+ | +------------+ | +--------+
^ | | ^
| V V |
| +-----------+ +-----------+ |
+---->| Relay-2 |<------------------->| Relay-3 |<---------+
P-2/P-4 +-----------+ +-----------+ P-2/P-4
Figure 6. Usage scenario for multipath transport
system framework in SIP System
(1) User A sends an INVITE to the SIP server that serves her domain.
(2) SIP server extracts the current addressable locations of the
caller and the callee, and the media information including media type
s and listed media codecs. Then SIP server sends an Openpath path
allocation request, ALLOCATE_PATH, carrying the above information to
the relay controller and requests it to assign the appropriate relay
paths for the media flow from user B to user A. In this example, the
relay controller selects two relay paths for the media flow from user
B to user A. They are (B, Relay-1, A) and (B, Relay-3, Relay-2, A),
named P-1 and P-2 respectively. And each relay path is associated
with a globally unique path identifier, named PID-1 and PID-2
respectively. The relay controller sends each of the three selected
relay server an ADD_PATH request message of OpenPath protocol.
ADD_PATH request includes the corresponding path identifier and
next-hop transport address of the receiving relay server. For
Relay-1, the path identifier is PID-1 and the next-hop transport
address is the current addressable location of user A. For Relay-2,
the path identifier is PID-2 and the next-hop transport address is
the current addressable location of user A. For Relay-3, the path
identifier is PID-2 and the next-hop transport address is Relay-2's
IP address and relay service port.
Leiwm, et al. Expires January 27, 2014 [Page 16]
Internet-Draft A Framework of Multipath Transport System July 2013
(3) SIP server inserts the path identifiers and the next-hop
transport addresses of user B into SDP body of INVITE and forward the
modified INVITE to user B. The next-hop transport addresses of user B
for P-1 and P-2 are separately the rely address of Relay-1 and
Relay-3.
(4) User B answers the call and sends back a 200 OK response.
(5) In the same way, SIP server obtains the allocated relay paths
for the media flow from user A to user B from the relay controller.
In this example, the relay controller assigns the same relay paths
with opposite direction for the media flow from user A to user B
based on symmetry principle. They are (A, Relay-1, B) and (A,
Relay-2, Relay-3, B), named P-3 and P-4 respectively, associated with
the path identifiers of PID-3 and PID-4. The relay controller sends
an ADD_PATH request message to each of the three selected relay
servers. For Relay-1, the path identifier is PID-3 and the next-hop
transport address is the current addressable location of user B. For
Relay-2, the path identifier is PID-4 and the next-hop transport
address is Relay-3's IP address and relay service port. For Relay-3,
the path identifier is PID-4 and the next-hop transport address is
the current addressable location of user B.
(6) SIP server inserts the path identifiers and the next-hop
transport addresses of user A into SDP body of 200 OK response and
forwards it to user A. The next-hop transport addresses of user A for
P-3 and P-4 are separately the rely address of Relay-1 and Relay-2.
Through the signaling process above, user A and user B separately
obtain three candidate paths including the default path and two relay
paths as the sending peer of the corresponding media flow. After
connectivity check, user A and user B can take advantage of multiple
paths to transport the media flow.
6. User Agent Behavior
Given multiple paths, user agent needs to provide essential support
for multipath transport, including session and path management, flow
partitioning and scheduling, subflow recombination, QoS feedback for
each subflow.
6.1 Multipath session management
In general, a session needs to be established between two
participants before transmitting data. At the same way, a multipath
session needs to be established before transporting data on multiple
paths. The multipath session setup uses a way of out-of-band
(e.g., SDP in SIP or RTSP). The capability exchange and relay path
Leiwm, et al. Expires January 27, 2014 [Page 17]
Internet-Draft A Framework of Multipath Transport System July 2013
advertisement should be done during the signaling process. A
multipath session may be setup from the beginning, or may be upgraded
from a normal single path session.
6.2 Path management
The user agent sender obtains the default path from the out-of-band
signaling for the session setup. In SIP/SDP case, the default path is
determined by the c= and m= lines in SDP. It may be the direct
network path between the sender and the receiver, or may traverse a
third-party node, such as a TURN server or a dedicated relay server.
For the latter, IP address and port of the third-party node is in the
c= and m= lines in SDP. The path identifier of the default path is
set to zero.
As stated in Section 5, the user agent sender can use two alternative
ways to collect candidate relay paths. One way is that the user agent
sender obtains candidate relay paths from the out-of-band signaling
for the session setup. The signaling server, which out-of-band
signaling passes through during session setup, requests the relay
controller to allocate relay paths for the session to be established
using OpenPath protocol. And then the signaling server inserts the
information of the allocated relay paths into corresponding signaling
messages to inform the participating user agents. Another way is that
the user agent sender obtains candidate paths through direct
interaction with the relay controller using OpenPath Protocol.
The user agent sender needs to perform path probes to determine if
the path is available and if so, obtains the transmission quality of
the path at the same time. After obtaining a new candidate path, the
user agent sender first performs path probe process, if the path is
available, marks it available and puts it into the available path
list ordered based on a decreasing priority level; if the path is not
available, marks it unavailable and puts it into the unavailable path
list. The user agent sender will periodically perform path probes for
all the paths in available path list to actively detect path failure
and perceive the dynamic changes to the path. If the probe process
for a particular path fails, the path will be marked unavailable and
removed from the available path list to the unavailable path list.
The user agent sender should also perform path probes for the paths
in unavailable path list in a longer cycle. If the probe process for
a particular path successes, the path will be marked available and
removed from the unavailable path list to the available path list.
If no data is received on a relay path within a given time, relay
servers will withdraw the corresponding resources allocated for this
path. Therefore, all the relay paths should be kept alive actively by
the user agent sender. To meet this requirement, the user agent
Leiwm, et al. Expires January 27, 2014 [Page 18]
Internet-Draft A Framework of Multipath Transport System July 2013
sender SHOULD send MPTP keep-alive packets periodically for both
active paths and non-active paths.
Using the information in the subflow MPTP reports, a user agent
sender can monitor delivery quality of active paths. If failure
(e.g., errors, unreachability, and congestion) happens in an active
path, the user agent sender may perform flow repartitioning and
spread the payload across other active paths, or may select a new
path from the available path list to replace the failure path.
6.3 Flow partitioning and scheduling
This document does not limit the usage of multiple paths. User agent
sender may concurrently use multiple paths to obtain higher
throughput, or may send all traffic on a specified path while all
other available paths are used only rarely to enhance resilience
(e.g., for retransmission, for error-repair, or for redundancy
packets). User agent sender MUST only use the default path and the
relay paths in available path list as the active path. How to select
multiple paths among all available paths and how many paths to use
concurrently are outside the scope of this document.
If multiple paths are used concurrently, the original data stream
should be divided into several substreams. Flow partitioning methods
are outside the scope of this document. Application-specific MPTP
documents may introduce some flow partitioning methods for specific
applications. A simple flow partitioning scheme is to assign packets
to multiple subflows using the round-robin algorithm. Specifically,
the original data stream is first divided into blocks of equal-sized
temporal or spacial length. Then, a subflow is assembled by picking
blocks periodically from the original blocks in an increasing order.
This method is simple and can be applied to all of the upper
applications. However, the data are treated equally and not
distinguished based on their different importance in terms of the
whole data flow. And great correlations exist among subflows which
are sent along paths with different transport capacity.
User agent sender should associate a subflow with an active path based
on a scheduling strategy. A scheduling policy should jointly consider
various factors including the estimated path bandwidth information,
the path reception statistics (e.g. RTT, loss-rate, delay etc.), the
relative importance of subflow data and so on. Due to the changes in
path characteristics, user agent sender should be able to change its
scheduling strategy during an ongoing session to fully explore the
potential of multipath transport.
Leiwm, et al. Expires January 27, 2014 [Page 19]
Internet-Draft A Framework of Multipath Transport System July 2013
6.4 Subflow packaging
In a multipath session, user agent sender formats data flow into MPTP
data packets following MPTP profile defined in this document and the
corresponding application-specific MPTP defined in one application-
specific MPTP documents. Then the user agent sender dispatches MPTP
data packets along corresponding relay paths.
The common header of MPTP packets includes two key fields: path
identifier and subflow-specific sequence number (SSSN). The path
identifier of a relay path, which is globally unique, is generated by
the relay controller. The path identifier of the default path is
fixed to zero. The path identifier is used to identify a specific
path by the user agent sender and relays. Meanwhile, the path
identifier is also used to identify a subflow by the user agent
receiver.
Subflow-specific sequence number is the sequence number of a MPTP
data packet in a specific subflow. It is used to help calculate the
quality of data delivery such as fractional losses, jitter, RTT, etc,
for each path. The initial subflow sequence number is randomly
generated when the subflow is first established in the multipath
session.
6.5 Subflow recombination
The user agent receiver recombines the original data flow according
to MPTP data packets received from multiple paths. The user agent
receiver first restores the order of each subflow using path
identifier and subflow sequence number in MPTP data packet headers.
Subsequent work is different according to specific applications.
Subsequent work SHOULD be defined in detail in application-specific
MPTP documents.
6.6 Subflow reporting
In order to monitor the quality of path delivery and to meet specific
requirements of some kind of applications, user agents SHOULD
generate MPTP reports for per subflow to provide subflow information.
The user agent sender generates MPTP Sender Reports (SR) for each
unique subflow and sends them along the same path as the subflow MPTP
data packets. As the relay path is unidirectional and the default
path is bidirectional, the user agent receiver generates MPTP
Receiver Reports (RR) for each unique subflow and sends them along
the default path.
Although subflow MPTP SRs and RRs are not sent along the same path,
they still can be used to measure the quality of path delivery. For
Leiwm, et al. Expires January 27, 2014 [Page 20]
Internet-Draft A Framework of Multipath Transport System July 2013
example, the calculated round-trip propagations to the user agent
receiver along different paths using subflow MPTP SRs and RRs still
can correctly represent relative size of transmission delays of
different paths.
When user agent generates MPTP report packets is outside the scope of
this document. It SHOULD be defined in detail in application-specific
MPTP documents. In general, timely MPTP reports are necessary for
multipath transport environments. MPTP reporting interval should be
frequent enough for user agents to quickly adapt to path fluctuation
and transmission errors. Meanwhile the traffic overhead introduced by
MPTP reporting SHOULD also be taken into account when designing an
application-specific MPTP for some specific applications.
In addition, what MPTP report packets contain and what to do when
receiving an MPTP report packet are related to the requirements of
specific application programs. They are outside the scope of this
document and SHOULD also be defined in detail in application-specific
MPTP documents.
7. Relay Behavior
Relay server performs MPTP packets lookups and forwarding based on a
local Path-Table. Relays are managed by the relay controller over
connections using a protocol referred to as OpenPath protocol.
7.1 Connection Management and Registrations
Relay server communicates with the relay controller over a connection
which may be encrypted using TLS or run directly over TCP. Relay
server initiates a standard TLS or TCP connection to the relay
controller. Relay server can get the IP address of the relay
controller in a number of ways, such as manual configuration, DNS
domain name queries and so on.
When a connection is established, relay server completes the
registration process by exchanging HELLO messages. The version field
in HELLO request is set to the highest OpenPath protocol version
supported by the relay server. Relay controller needs to check the
included OpenPath protocol version in HELLO request. If the version
is not supported, relay controller responds to the HELLO with a
failure response.
The relay server identifier field is set to zero in the initial HELLO
request. This indicates that it is the first time for this relay
server to register with the relay controller. The relay controller
needs to assign a unique identifier for this relay server and insert
the assigned relay server identifier in the corresponding HELLO
Leiwm, et al. Expires January 27, 2014 [Page 21]
Internet-Draft A Framework of Multipath Transport System July 2013
response message. Subsequent HELLO request messages may carry the
relay server identifier directly.
The HELLO request contains the IP address and port of the relay
server for the relay service.
If a failure response to the HELLO message is received, relay server
then terminates the connection. Otherwise, the connection proceeds
and the relay server may start relay service.
During the lifetime of the connection, ECHO requests should be sent
periodically from either relay server or relay controller, and the
request receiver must return an ECHO reply.
After connection establishment, relay server may receive FEATURES
requests from relay controller. It must respond with a FEATURES reply
that specifies its capabilities.
During the lifetime of the connection, relay controller may
periodically collect statistics from a relay server by STATISTICS
requests. The relay server must respond with a STATISTICS reply that
specifies its current statistics.
7.2 Path-Table Management
The Path-Table contains a set of path entries each of which
corresponds to an associated relay path. Each path entry consists of
match fields, result fields and counters (see table 1).
Match fields are used to match against MPTP packets. Match fields
only consist of path identifier in this document.
Path Identifier: a 32 bit unsigned integer uniquely identifying the
associated relay path.
Result fields include next-hop transport address, idle timeout and
hard timeout. Next-hop transport address is to determine where the
matched packets are forwarded. Idle timeout and hard timeout are used
for relay to actively clear those expired path entries.
Next-hop Transport Address Type: corresponds to the type of the
next-hop transport address. Namely:
1: IPv4 address
2: IPv6 address
Next-hop Transport IP Address: the address to which the relay server
forwards the matched packets.
Leiwm, et al. Expires January 27, 2014 [Page 22]
Internet-Draft A Framework of Multipath Transport System July 2013
Next-hop Transport Port: the port number to which the relay server
forwards the matched packets.
Counters are maintained for each path entry and updated for matching
MPTP packets.
Received packets: the amount of packets the path has matched.
Received bytes: the amount of bytes the path has matched.
Duration: the amount of time the path has been installed in the relay
server.
Table 1: Fields in a path entry.
+-------------------------------------+---------------------------+
| Fields | Bits |
+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+
| Match fields |
+-----------------------------------------------------------------+
| Path Identifier | 32 |
+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+
| Result fields |
+-----------------------------------------------------------------+
| Next-hop transport address type | 8 |
+-------------------------------------+---------------------------+
| | For an IPv6 address, this |
|Next-hop transport IP address | is 128;for an IPv4 address|
| | , this is 32. |
+-------------------------------------+---------------------------+
| Next-hop transport port | 16 |
+-------------------------------------+---------------------------+
| Idle timeout | 16 |
+-------------------------------------+---------------------------+
| Hard timeout | 16 |
+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+
| Counters |
+-----------------------------------------------------------------+
| Received packets | 64 |
+-------------------------------------+---------------------------+
| Received bytes | 64 |
+-------------------------------------+---------------------------+
| Duration (seconds) | 32 |
+-------------------------------------+---------------------------+
| Duration (nanoseconds | 32 |
+-----------------------------------------------------------------+
Path-Table of relay server is managed by relay controller through
Leiwm, et al. Expires January 27, 2014 [Page 23]
Internet-Draft A Framework of Multipath Transport System July 2013
Path-Table modification messages. For ADD_PATH requests, relay server
must first check if any path entry with the same path identifier has
existed in the Path-Table. If an overlap conflict exists between an
existing path entry and the ADD_PATH request, relay server must
refuse the addition and respond with a failure response. For valid
ADD_PATH requests, relay server must insert a new path entry in the
Path-Table and respond to the relay controller with a success
response. The counters of received packets and received bytes in the
new inserted entry are set to zero.
For DELETE_PATH requests, if a matching entry exists in the Path-
Table,it must be removed, and a success response is returned to the
relay controller. For DELETE_PATH requests, if no path entry matches,
no path entry modification occurs, and a failure response is returned
to the relay controller.
For UPDATE_PATH requests, if a matching entry exists in the
Path-Table, the result fields of this entry is updated with the value
from the request, and counter fields are left unchanged. For
UPDATE_PATH requests, if no path entry matches, no path entry
modification occurs, and a failure response is returned to the relay
controller.
7.3 Path Validity Management
Each path entry has an idle timeout and a hard timeout associated
with it. These two fields control how quickly a path entry expires.
When a path entry is inserted by an ADD_PATH request or modified by a
UPDATE_PATH request, its idle timeout and hard timeout are set with
the values from the message.
If either value is non-zero, relay server must note the arrival time
of MPTP packet on the associated path, as it may need to evict the
path entry later. A non-zero hard timeout field causes the path entry
to be removed after the given number of seconds, regardless of how
many ackets it has matched. A non-zero idle timeout field causes the
path entry to be removed when it has matched no packets in the given
number of seconds. Using these two fields, relay server can actively
clear those expired path entries. In addition, relay controller may
actively remove path entries by sending DELETE_PATH messages.
If a relay server actively removes a path entry, it must send a
NOTIFY message to relay controller. The NOTIFY message contains a
complete description of the path entry including the reason for
removal, the path entry duration and statistics at the time of
removal.
Leiwm, et al. Expires January 27, 2014 [Page 24]
Internet-Draft A Framework of Multipath Transport System July 2013
7.4 Relay Service Management
After successful registration, relay server may send a START message
to relay controller to indicate that it has enough capacity to
provide relay service.
In the case that a relay server is overloaded, or under other some
situations, the relay server may send a STOP message to relay
controller to indicate that it will no longer receive new relay
service requests (i.e. ADD_PATH messages) for a while. During this
period, the relay server still provides relay service for those
existing relay paths. And ECHO messages still need to be sent
periodically between the relay server and the relay controller.
When a relay server recovers from an overloaded state, it may send a
START message to relay controller to indicate that it has additional
capacities to provide new relay services.
When a relay server wants to stop relay service permanently, it
should actively send a BYE message to relay controller before
terminating the connection. In this way, relay controller can be
ready in time to do some remedial measures. For instance, relay
controller may assign new relay paths for the affected media flow.
7.5 MPTP Packet Processing
The main function of a relay server is to provide relay service to
the associated subflows. All subflows associated with a relay server
share a common relay port of this relay server.
After receiving a MPTP packet, relay server extracts the path
identifier from the received MPTP packet and does matching in the
Path-Table. If the received MPTP packet does not match a path entry
in the Path-Table, relay server has nothing to do but to drop the
packet. If the received packet matches a path entry in the
Path-Table, relay server forwards it to the associated next-hop
transport address in the matched path entry. Meanwhile, relay server
updates the associated statistical counters in the matched path
entry.
8. Relay Controller Behavior
Relay controller is responsible for managing relay servers and
selecting one or multiple relay paths for a data flow.
8.1 Relay Server Management
Relay controller manages the topology of a network of relay servers
Leiwm, et al. Expires January 27, 2014 [Page 25]
Internet-Draft A Framework of Multipath Transport System July 2013
in its region, and maintains registration and status information of
relay servers such as relay capacity, availability and work load. For
each relay server, relay controller collects its capabilities
information by FEATURES messages, and periodically collects its
statistics information by STATISTICS messages.
8.2 Relay Path Allocation
During establishing a multipath session, the user agent sender or the
signaling server which out-of-band signaling pass through during
session setup, requests relay controller to allocate relay paths by
ALLOCATE_PATH messages. The ALLOCATE_PATH request carries the related
information of the data flow which is going to establish, such as
location information of participants, transport requirements of the
data flow and so on. Relay controller selects one or multiple relay
paths according to the information carried in ALLOCATE_PATH request
and the status information of the registered relay servers. For each
selected relay path, the relay controller sends an ADD_PATH request
to each relay server on the selected relay path. The ADD_PATH request
carries the path identifier, next-hop transport address of the
receiving relay server, etc. A relay path is assigned successfully
only when each relay server on this relay path replies with an
ADD_PATH success response. After successful path allocation, relay
controller replies an ALLOCATE_PATH success response to inform the
user agent sender or the signaling server about the allocated relay
path information including the path identifier, user agent sender's
next-hop transport address, etc.
9. OpenPath Protocol
9.1 Protocol Overview
OpenPath is based on a request/response transaction model. Each
transaction consists of a request and a response. A response uses the
same transaction id as is in the associated request to facilitate
pairing. The transaction id is a 32-bit identifier generated by the
request sender.
OpenPath messages are guaranteed delivery over a connection-oriented
channel. All integer fields are carried in network octet order, that
is, most significant byte first. Octets designated as padding have
the value zero.
All OpenPath messages begin with an OpenPath common header. OpenPath
messages MAY contain a message body. The structure and interpretation
of a body depends on the message type.
OpenPath message types fall into four classes: relay-to-controller,
Leiwm, et al. Expires January 27, 2014 [Page 26]
Internet-Draft A Framework of Multipath Transport System July 2013
controller-to-relay, user agent-to-controller and symmetric.
Relay-to-controller messages are initiated by relay server and used
to manage the channel connection and update relay controller of
changes to the relay server. Controller-to-relay messages are
initiated by relay controller and used to manage the Path-Table or
inspect the state of relay servers. User agent-to-controller messages
are initiated by user agent or out-of-band signaling server, and used
for relay path allocation. Symmetric messages are initiated by either
relay controller or relay server.
9.1.1 Relay-to-Controller
Relay-to-controller message is initiated by a relay server and
requires a response message from relay controller.
HELLO: Relay server registers with relay controller by sending a
HELLO request. Relay controller must respond with a HELLO response
that indicates the outcome of registration. This is commonly
performed upon establishment of the OpenPath connection channel.
The relay server identifier field is set to zero in the initial HELLO
request. Relay controller must assign a unique identifier for this
relay server and insert the assigned relay server identifier in the
corresponding HELLO response. Subsequent HELLO requests may carry the
assigned relay server identifier directly.
The HELLO request contains a message body. The body contains the IP
address and port of relay server for the relay service.
START: Relay server starts relay service by sending a START request.
Relay controller must respond with a START response that indicates
the outcome of relay service startup.
STOP: Relay server may stop relay service temporarily by sending a
STOP request. For instance, when a relay server is overloaded, it may
want to refuse accepting any new relay service requests for a while.
Relay controller must respond with a STOP response that indicates the
outcome of relay service pause. During relay service pause, this
relay server still provides relay service for those existing relay
paths. This relay server may restart relay service by sending a START
request after a while.
BYE: Relay server may terminate the relay service permanently by
sending a BYE request, such as before terminating the OpenPath
connection channel or exiting. Meanwhile, this relay server ceases
relay service for all existing relay paths. If this relay server
wants to start relay service again, it must first perform
Leiwm, et al. Expires January 27, 2014 [Page 27]
Internet-Draft A Framework of Multipath Transport System July 2013
registration with relay controller.
NOTIFY: When a relay server actively removes a path entry, it may
notify relay controller by sending a NOTIFY request. For instance, in
order to save the resource, relay server actively removes those path
entries which lack activity.
9.1.2 Controller-to-Relay
Controller-to-relay message is initiated by a relay controller and
require a response message from relay server.
FEATURES: Relay controller may request the capabilities of a relay
server by sending a FEATURES request. This relay server must respond
with a FEATURES response that specifies its capabilities. This is
commonly performed after successful registration of this relay
server.
STATISTICS: Relay controller may periodically collect statistics from
relay servers by sending STATISTICS requests. Relay server must
respond with a STATISTICS response that specifies its current
statistics.
ADD_PATH: When relay controller assigns relay paths for a data flow,
it must send an ADD_PATH request to each relay server on the assigned
relay path. The relay server must respond with an ADD_PATH response
that specifies the outcome of adding a new path entry. A relay path
is assigned successfully only when each relay server replies with an
ADD_PATH success response.
UPDATE_PATH: Relay controller may want to modify an existing path
entry in the Path-Table of a relay server by sending a UPDATE_PATH
request. For instance, a data flow may be "put on hold" and data
transmission may be ceased for a while. In this case, relay
controller may update the idle timeout and hard timeout of the
corresponding path entries of all the relay servers on the affected
relay paths to a longer time. Relay server must respond with an
UPDATE_PATH response that specifies the outcome of updating an
existing path entry.
DELETE_PATH: Relay controller may delete an existing path entry in
the Path-Table of a relay server by sending a DELETE_PATH request,
such as when a data flow ends normally. Relay server must respond
with a DELETE_PATH response that specifies the outcome of deleting an
existing path entry.
Leiwm, et al. Expires January 27, 2014 [Page 28]
Internet-Draft A Framework of Multipath Transport System July 2013
9.1.3 User agent-to-Controller
User agent-to-controller message is initiated by a user agent or an
out-of-band signaling server, and requires a response message from
relay controller.
ALLOCATE_PATH: During establishing a multipath session, user agent
sender may request relay controller to allocate candidate relay paths
by exchanging ALLOCATE_PATH messages directly with relay controller.
Or the signaling server requests relay controller to allocate
candidate relay paths for user agents, and inserts the information of
the allocated relay paths into corresponding signaling messages to
inform user agents. Relay controller must respond with an
ALLOCATE_PATH response that specifies the outcome of path allocation
and the information of the assigned relay path if exists.
9.1.4 Symmetric
Symmetric message is sent in either direction between relay server
and relay controller.
ECHO: ECHO requests MUST be sent periodically from either relay
controller or relay server. ECHO messages can be used to measure the
latency of the connection between relay controller and relay server,
as well as verify the peer's liveness. The receiver of an ECHO request
must respond with an ECHO response, which consists of an OpenPath
header plus the unmodified body of an ECHO request message.
9.2 Common Structures
This section describes several common structures used by multiple
messages.
9.2.1 OpenPath Common Header
All OpenPath messages begin with an OpenPath common header which is
defined below:
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| V |R|S| Type | Length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Peer Identifier |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Transaction Identifier |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Leiwm, et al. Expires January 27, 2014 [Page 29]
Internet-Draft A Framework of Multipath Transport System July 2013
This field identifies the version of OpenPath protocol version being
used. The version defined by this document is one (1).
R: 1 bit
If the R bit is set, the message is an OpenPath request; otherwise,
the message is an OpenPath response.
S: 1 bit
When the message is a request, this bit is reserved. When the message
is a response, if the S bit is set, the message is a success
response; otherwise, the message is a failure response.
Type: 8 bits
This field identifies the type of the OpenPath messages. This
document defines 13 OpenPath message types:
+----------------------------------------------------------------+
| Type Value | Type Name | Sending Direction |
+--------------+---------------+---------------------------------+
| 1 | HELLO | RTP relay -> Controller |
+--------------+---------------+---------------------------------+
| 2 | BYTE | RTP relay -> Controller |
+--------------+---------------+---------------------------------+
| 3 | ECHO | RTP relay -> Controller Or |
| | | Controller -> Relay |
+--------------+---------------+---------------------------------+
| 4 | START | RTP relay -> Controller |
+--------------+---------------+---------------------------------+
| 5 | STOP | RTP relay -> Controller |
+--------------+---------------+---------------------------------+
| 6 | NOTIFY | RTP relay -> Controller |
+--------------+---------------+---------------------------------+
| 7 | FEATURES | Controller -> Relay |
+--------------+---------------+---------------------------------+
| 8 | STATISTICS | Controller -> Relay |
+--------------+---------------+---------------------------------+
| 9 | ADD-PATH | Controller -> Relay |
+--------------+---------------+---------------------------------+
| 10 | DETELE-PATH | Controller -> Relay |
+--------------+---------------+---------------------------------+
| 11 | UPDATE-PATH | Controller -> Relay |
+----------------------------------------------------------------+
| 12 | ALLOCATE_PATH | User Agent -> Controller |
+--------------+---------------+---------------------------------+
| 13 | RELEASE_PATH | User Agent -> Controller |
+----------------------------------------------------------------+
Length: 16 bits
Leiwm, et al. Expires January 27, 2014 [Page 30]
Internet-Draft A Framework of Multipath Transport System July 2013
The length field indicates the total length of the message in 32-bit
words including the header and any padding.
Peer Identifier: 32 bits
This field is a 32-bit identifier that corresponds to a globally
unique relay server, user agent or out-of-band signaling server. It
is generated by relay controller during registration process of a
relay server, user agent or out-of-band signaling server. This
identifier remains unchanged during the entire lifecycle of the
OpenPath connection with relay controller. For OpenPath messages
exchanged between relay controller and relay server, this field
corresponds to a relay server identifier. For OpenPath messages
exchanged between relay controller and user agent, this field
corresponds to a user agent identifier. For OpenPath messages
exchanged between relay controller and out-of-band signaling server,
this field corresponds to a signaling server identifier. For the last
two cases, it is uniformly called user agent identifier.
Transaction Identifier: 32 bits
This field identifies the transaction id associated with this
message. It is randomly generated by the request sender and discarded
when the associated response message is received. The transaction
identifiers of responses must always match the requests that prompted
them.
9.2.2 Common Body of OpenPath Failure Responses
OpenPath failure responses of all message types MAY contain an
optional body after the OpenPath common header. If present, the body
conforms to a common structure defined below.
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Status code | Rlength | reason |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| reason |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Status code: 8 bits
This field is a numeric result code that indicates the outcome of a
request processing.
Rlength: 8 bits
This field is the length of the reason phrase in 16-bit word length.
Reason: variable size
This field is a short textual description of the status code.
Leiwm, et al. Expires January 27, 2014 [Page 31]
Internet-Draft A Framework of Multipath Transport System July 2013
9.2.3 Transport Address Structure
Relay server needs to advertise relay controller about its transport
address for relay service. Relay controller also needs to tell relay
server about the transport address of its next-hop node for each
associated path. A transport address is defined as follow.
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Address Type | Pad(0) | Port |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Address |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Address Type: 8 bits
This field is the type of transport address. Namely:
1: IPv4 address
2: IPv6 address
Port: 16 bits
This field is the port number part of transport address.
Address: 4 octets (IPv4), 16 octets (IPv6)
The IP address part of transport address.
9.3 Message Format of OpenPath Request and Success Response
9.3.1 HELLO
A HELLO request contains a body beyond an OpenPath common header.
The body contains the transport address of the relay server to
provide relay service.
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| V |R|S| Type | Length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Relay Server Identifier |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Transaction Identifier |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Address Type | Pad(0) | Port |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Address |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Leiwm, et al. Expires January 27, 2014 [Page 32]
Internet-Draft A Framework of Multipath Transport System July 2013
A HELLO success response only contains an OpenPath common header.
9.3.2 START/STOP/BYE
START/STOP/BYE requests and their success responses have no body;
that is, they only contain an OpenPath common header.
9.3.3 ECHO
An ECHO request consists of an OpenPath common header and an optional
body. If the body is absent, an ECHO transaction is used to simply
verify liveness between relay controller and relay server. If the
body is present, it may contain a timestamp field to check latency
between relay controller and relay server. Example:
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| V |R|S| Type | Length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Relay Server Identifier |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Transaction Identifier |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| NTP timestamp, most significant word |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| NTP timestamp,least significant word |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
NTP timestamp: Using the timestamp format of the Network Time
Protocol (NTP), which is in seconds relative to 0h UTC on 1 January
1900 [2]. The full resolution NTP timestamp is a 64-bit unsigned
fixed-point number with the integer part in the first 32 bits and the
fractional part in the last 32 bits.
An ECHO response has the same message format as the associated ECHO
request.
9.3.4 NOTIFY/DELETE_PATH
NOTIFY/DELETE_PATH requests contain a body beyond an OpenPath common
header. The body only contains a path identifier field.
Leiwm, et al. Expires January 27, 2014 [Page 33]
Internet-Draft A Framework of Multipath Transport System July 2013
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| V |R|S| Type | Length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Relay Server Identifier |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Transaction Identifier |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Path Identifier |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Path Identifier: 32 bits
This field is a 32-bit identifier that corresponds to a globally
unique path. It is generated by relay controller when assigning relay
paths for a data flow.
NOTIFY/DELETE_PATH success responses only contain an OpenPath common
header.
9.3.5 ADD-PATH/ UPDATE_PATH
ADD-PATH/UPDATE_PATH requests contain a body beyond an OpenPath
common header. The body contains match fields and result fields of a
path entry. The match fields contain a single path identifier field.
The result fields contain next-hop transport address and idle/hard
timeout fields.
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| V |R|S| Type | Length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Relay Server Identifier |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Transaction Identifier |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Path Identifier |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Address Type | Pad(0) | Port |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Address |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Idle timeout | Hard timeout |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
Idle timeout: 16 bits
This field is the number of seconds after which the path will timeout
Leiwm, et al. Expires January 27, 2014 [Page 34]
Internet-Draft A Framework of Multipath Transport System July 2013
if no traffic.
Hard timeout: 16 bits
This field is the number of seconds after which the path must expire
regardless of whether or not packets go through the path.
ADD-PATH/ UPDATE_PATH success responses only contain an OpenPath
common header.
9.3.6 ALLOCATE_PATH/ RELEASE_PATH
(to be continued)
9.3.7 FEATURES
A FEATURES request only contains an OpenPath common header.
A FEATURES success response contains a body beyond an OpenPath common
header.
(to be continued)
9.3.8 STATISTICS
A STATISTICS request only contains an OpenPath common header.
A STATISTICS success response contains a body beyond an OpenPath
common header.
(to be continued)
10. MPTP Profile
10.1 Overview
As stated in Section 5, MPTP obtain a simple, unreliable datagram
service from the underlying UDP. MPTP SHOULD meet various delivery
requirements of upper applications. As stated in the introduction, in
order to be extensible and suitable for various applications, MPTP is
designed to be a protocol suite which consists of one MPTP profile
and multiple application-specific MPTPs. The MPTP profile only
defines common rules of multipath transport based on
application-level relay for various upper applications. It is aimed
to provide application-level multipath routing mechanism. Other
transmission requirements of upper applications, such as timely or
reliable delivery, SHOULD be met in corresponding
application-specific MPTP documents.
In addition to carrying payload data passed from upper application
programs through multiple paths, MPTP also need to provide path
control functions including keeping path alive and monitoring the
Leiwm, et al. Expires January 27, 2014 [Page 35]
Internet-Draft A Framework of Multipath Transport System July 2013
quality of data delivery on each path. Therefore, MPTP packets are
divided into two types: MPTP data packets and MPTP control packets.
MPTP control packets include MPTP keep-alive packets and MPTP report
packets.
User agent sender formats the payload data passed from upper
application programs into MPTP data packets which are sent along the
associated path.
User agent sender SHOULD send MPTP keep-alive packets periodically
for each path including both active path and non-active path.
To monitor the transport quality of a path, user agents generate MPTP
report packets for each subflow. Due to the content of MPTP report
packets depends on the delivery requirements of upper application
programs, this document does not give concrete content and processing
of the MPTP report packets, which should be described in
application-specific MPTP documents. Here, we only outline the
delivery methods of MPTP report packets. The user agent sender
generates MPTP sender reports for each subflow and sends them along
the same path as the subflow MPTP data packets. The user agent
receiver responds with MPTP receiver reports which are sent along the
default path.
10.2 MPTP Fixed Header Fields
All MPTP packets have a fixed eight octet-length header. The first
four octets except for the first four bits correspond to different
fields for MPTP data packets and MPTP control packets.
10.2.1 Fixed Header Fields of MPTP Data Packet
Fixed header of MPTP data packet is defined below:
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|V=1|T|P| AMT | TOS | SSSN |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Path Identifier |
+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+
| |
| Payload Data |
| |
| .... |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
The fields have the following meaning:
Leiwm, et al. Expires January 27, 2014 [Page 36]
Internet-Draft A Framework of Multipath Transport System July 2013
version (V): 2 bits
This field identifies the version of MPTP. The version defined by
this document is one (1).
MPTP packet type (T): 1 bit
This field identifies the type of a MPTP packet. If the MPTP packet
type bit is set, this packet is a MPTP data packet; otherwise, this
packet is a MPTP control packet.
padding (P): 1 bit
If the padding bit is set, the packet contains one or more additional
padding octets at the end which are not part of the payload. The last
octet of the padding contains a count of how many padding octets
should be ignored, including itself. Padding may be needed by some
encryption algorithms with fixed block sizes.
Application-specific MPTP type (AMT): 4 bits
This field identifies the type of application-specific MPTP that this
packet format conforms to. This document defines two
application-specific MPTP types: AMT = 1, indicates that this MPTP
packet conforms to RTP-based multimedia application-specific MPTP;
AMT = 2, indicates that this MPTP packet conforms to reliable
application-specific MPTP.
type of service (TOS): 8 bits
The last four bits of the type of service is reserved. The first four
bits of the type of service, similar to TOS field in internet packet
header, is specified along the abstract parameters precedence, delay,
throughput, and reliability. These abstract parameters embody the
delivery requirements of upper application programs. The values of
these abstract parameters should be specified in application-specific
MPTP documents.
Precedence: An independent measure of the importance of the
payload data.
Delay: Prompt delivery is important for the payload data with
this indication.
Throughput: High data rate is important for the payload data with
this indication.
Reliability: A higher level of effort to ensure delivery is
important for the payload data with this indication.
Subflow-Specific Sequence Number (SSSN):
This field is the sequence of the MPTP data packet in the subflow.
Each subflow has its own strictly monotonically increasing sequence
Leiwm, et al. Expires January 27, 2014 [Page 37]
Internet-Draft A Framework of Multipath Transport System July 2013
number space.
Path Identifier:
This field is the identifier of the path that the MPTP packet is
associated with. For a relay path, the path identifier is globally
unique; for the default path, the path identifier is fixed to zero.
Payload Data: variable size
The content of payload data should be described in
application-specific MPTP documents.
10.2.2 Fixed Header Fields of MPTP Control Packet
Fixed header of MPTP control packet is defined below:
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|V=1|T|P| MET | CT | Length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Path Identifier |
+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+
| |
| Control Data |
| |
| .... |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
The fields of V, T, P, MET and Path Identifier have the same meaning
as those fields in MPTP data packet. Other fields have the following
meaning:
MPTP control packet type (CT): 8 bits
This field identifies the type of MPTP control packet. Namely:
Leiwm, et al. Expires January 27, 2014 [Page 38]
Internet-Draft A Framework of Multipath Transport System July 2013
+-------------------------------------------------------------------+
|MPTP control packet type (CT) |Use |
+---------------------------------+---------------------------------+
|1 |Subflow Sender Report, for |
| |transmission statistics from the |
| |user agent sender |
+---------------------------------+---------------------------------+
|2 |Subflow Receiver Report, for |
| |reception statistics from the |
| |user agent receive |
+---------------------------------+---------------------------------+
|3 |Keep Alive, for keeping a path |
| |alive |
+-------------------------------------------------------------------+
Length: 8 bits
This field is the length of the encapsulated control data after the
MPTP fixed header in byte length.
Control Data: variable size
For MPTP keep-alive packet, control data may be empty.
For MPTP report packet, the content of control data should be
described in application-specific MPTP documents.
11. SDP Considerations
The advertisement of MPTP capability and relay paths MUST be
performed by out-of-band signaling, for example, as part of a SIP
offer/answer exchange using SDP. This section defines dedicated
extensions in SDP. SDP syntax and procedures are available that can
be re-used or easily adapted to provide the necessary capabilities.
11.1 Signaling MPTP Capability in SDP
A communication participant, who follows the framework of multipath
transport system based on application-level relay, MUST use SDP to
indicate that it supports and wants to use this framework. The
mptp-relay attribute defined here will be used to indicate the
support for this framework. The mptp-relay attribute is a media level
parameter. The syntax of the mptp-relay attribute is defined using
the following Augmented BNF, as defined in [RFC5234].
mptp-relay-attrib = "a=" "mptp-relay" ":" mptp-name CRLF
The mptp-name field indicates an application-specific MPTP. In an
application-specific MPTP document, the value of mptp-name field MUST
be given for that application-specific MPTP.
When SDP is used following the offer/answer mode [RFC3264], the
Leiwm, et al. Expires January 27, 2014 [Page 39]
Internet-Draft A Framework of Multipath Transport System July 2013
"mptp-relay" attribute indicates the desire to transport media flow
on multiple paths. If the offerer supports and wishes to use MPTP,
it MUST include a media-level "a=mptp-relay" attribution in the
initial SDP offer. If the initial SDP offer contains "a=mptp-relay"
attribution and if the answerer supports and wishes to use MPTP, it
MUST include this attribute in the SDP answer.
When SDP is used in a declarative manner, the "mptp-relay" attribute
indicates that the message sender can send or receive media data over
multiple paths. The message receiver SHOULD be capable to stream
media to multiple paths or be prepared to receive media from multiple
paths.
11.2 Relay Path Advertisement in SDP
The information of candidate relay paths MUST be advertised to the
user agent sender in the out-of-band signaling. The relay-path
attribute is extended to advertise candidate relay paths in SDP. The
syntax of the relay-path attribute is defined using the following
Augmented BNF, as defined in [RFC5234]. The definitions of DIGIT, SP
and CRLF are according to RFC4234.
relay-path-attrib = "a=" relay-path-label ":" counter SP path-id SP
transport-address CRLF
relay-path-label = "relay-path"
counter = 1*DIGIT
path-id = 32token-char
transport-address = addrtype ":" unicast-address ":" port
; addrtype from RFC4566, typically "IP4" | "IP6"
; port from RFC4566
unicast-address = IP4-address / IP6-address
; IP4-address from RFC4566
; IP6-address from RFC4566
token-char = %x21 / %x23-27 / %x2A-2B / %x2D-2E / %x30-39 / %x41-5A /
%x5E-7E
The path identifier and the next-hop transport address of the user
agent sender for each candidate relay path is encapsulated in a
relay-path attribute.
The <counter> parameter indicates the priority of the associated
relay path and it is a monotonically increasing positive integer
starting at 1. Number 1 is the highest priority. The counter must be
reset for each media line. The relay-path attributes are ordered
based on a decreasing priority level.
The <path-id> parameter indicates the path identifier of the
associated relay path.
Leiwm, et al. Expires January 27, 2014 [Page 40]
Internet-Draft A Framework of Multipath Transport System July 2013
The <transport-address> is the next-hop transport address of the
user agent sender for associated candidate relay path. The <addrtype>
is the address type. This document only defines IP4 and IP6 for IPv4
addresses and IPv6 addresses respectively.
Example:
a=relay-path:1 1a3b6c9d0e2f6g1qazxsw23edcvfr45t IP4:192.0.3.2:10000
12. IANA Considerations
The following contact information shall be used for all registrations
in this document:
Contact: Weimin Lei
mailto:leiweimin@ise.neu.edu.cn
tel:+86-24-8368-3048
Note to the RFC-Editor: When publishing this document as an RFC,
please replace "RFC XXXX" with the actual RFC number of this document
and delete this sentence.
12.1 SDP Attributes
In the registry for SDP parameters, the attributes named "mptp-relay"
and "relay-path" have been registered as follows:
SDP Attribute ("att-field"):
Attribute Name: mptp-relay
Long form: Multipath transport system framework based on
application-level relay
Type of name: att-field
Type of attribute: Media level only
Subject to charset: No
Purpose: This attribute is used to indicate support for
multipath transport system framework based on
application-level relay.
Reference: See this document
Values: See this document.
SDP Attribute ("att-field"):
Attribute Name: relay-path
Long form: Relay Path of multipath transport system
framework based on application-level relay
Type of name: att-field
Type of attribute: Media level only
Leiwm, et al. Expires January 27, 2014 [Page 41]
Internet-Draft A Framework of Multipath Transport System July 2013
Subject to charset: No
Purpose: This attribute is used to describe the
information of a relay path including the path
identifier and the next-hop transport address
of the user agent sender.
Reference: See this document
Values: See this document.
13. Security Considerations
TBD
14. References
14.1 Normative References
[1] Bradner, S., "Key words for use in RFCs to Indicate Requirement
Levels", BCP 14, RFC 2119, March 1997.
[2] Mills, D., "Network Time Protocol (Version 3) Specification,
Implementation and Analysis", RFC 1305, March 1992.
14.2 Informative References
[3] Schulzrinne, H., Casner, S., Frederick, R., and V. Jacobson,
"RTP: A Transport Protocol for Real-Time Applications", STD 64,
RFC 3550, July 2003.
[4] Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, A.,
Peterson, J., Sparks, R., Handley, M., and E. Schooler, "SIP:
Session Initiation Protocol", RFC 3261, June 2002.
[5] Handley, M., Jacobson, V., and C. Perkins, "SDP: Session
Description Protocol", RFC 4566, July 2006.
[6] Schulzrinne, H., Rao, A., Lanphier, R., "Real Time Streaming
Protocol (RTSP)", RFC2326, April 1998.
Authors' Addresses
Weimin Lei
Northeastern University
Institute of Communication and Information System
College of Information Science and Engineering
Shenyang, China, 110819
P. R. China
Phone: +86-24-8368-3048
Leiwm, et al. Expires January 27, 2014 [Page 42]
Internet-Draft A Framework of Multipath Transport System July 2013
Email: leiweimin@ise.neu.edu.cn
Wei Zhang
Northeastern University
Institute of Communication and Information System
College of Information Science and Engineering
Shenyang, China, 110819
P. R. China
Email: zhangwei1@ise.neu.edu.cn
Shaowei Liu
Northeastern University
Institute of Communication and Information System
College of Information Science and Engineering
Shenyang, China, 110819
P. R. China
Email: liushaowei.ise.neu@gmail.com
liu_nongfu@163.com
Leiwm, et al. Expires January 21, 2014 [Page 43]