Network Working Group Allison Mankin
INTERNET-DRAFT USC/ISI
Allyn Romanow
MCI
Scott Bradner
Harvard University
Vern Paxson
LBL
With the TSV
Area Directorate
May 1998
IETF Criteria for Evaluating Reliable Multicast Transport
and Application Protocols
<draft-mankin-reliable-multicast-02.txt>
Status of this Memo
This document is an Internet Draft. Internet Drafts are working
documents of the Internet Engineering Task Force (IETF), its Areas,
and its Working Groups. Note that other groups may also distribute
working documents as Internet
Internet Drafts are draft documents valid for a maximum of six
months. Internet Drafts may be updated, replaced, or obsoleted by
other documents at any time. It is not appropriate to use Internet
Drafts as reference material or to cite them other than as a "working
draft" or "work in progress."
To learn the current status of any Internet-Draft, please check the
"1id-abstracts.txt" listing contained in the internet-drafts Shadow
Directories on: ftp.is.co.za (Africa) nic.nordu.net (Europe)
ds.internic.net (US East Coast) ftp.isi.edu (US West Coast)
munnari.oz.au (Pacific Rim)
Abstract
This memo describes the procedures and criteria for reviewing
reliable multicast protocols within the Transport Area (TSV) of the
IETF. Within today's Internet, important applications exist for a
reliable multicast service. Some examples that are driving reliable
multicast technology are collaborative workspaces (such as
whiteboard), data and software distribution, and (more speculatively)
web caching protocols. Due to the nature of the technical issues, a
single commonly accepted technical solution that solves all the
demands for reliable multicast is likely to be infeasible [RMMinutes
1997].
A number of reliable multicast protocols have already been developed
Mankin/Romanow/Bradner/Paxson [Page 1]
Internet-Draft Evaluating Reliable Multicast May 1998
to solve a variety of problems for various types of applications.
[Floyd97] describes one widely deployed example. How should these
protocols be treated within the IETF and how should the IETF guide
the development of reliable multicast in a direction beneficial for
the general Internet?
The TSV Area Directors and their Directorate have outlined a set of
review procedures that address these questions and set criteria and
processes for the publication as RFCs of Internet-Drafts on reliable
multicast transport protocols.
1.0 Background on IETF Processes and Procedures
In the IETF, work in an area is directed and managed by the Area
Directors (ADs), who have authority over the chartering of working
groups (WGs).
In addition, ADs review individually submitted (not by WGs) Internet-
Drafts about work that is relevant to their areas prior to
publication as RFCs (Experimental, Informational or, in rare cases,
Standards Track). The review is done according to the guidelines set
out in the Internet Standards Process, RFC 2026 [InetStdProc96].
The purpose of this document is to present the criteria that will be
used by the TSV ADs in reviewing reliable multicast Internet-Drafts
for Standards Track RFCs.
2.0 Introduction
There is a strong application demand for reliable multicast.
Widespread use of the Internet makes the economy of multicast
transport attractive. The current Internet multicast model offers
best-effort many-to-many delivery service and offers no guarantees.
Reliable multicast transports add delivery guarantees, not
necessarily like those of reliable unicast TCP, to the group-delivery
model of multicast. A panel of some major users of the Internet,
convened at the 38th IETF, articulated reliable bulk transfer
multicast as one of their most critical requirements [DiffServBOF97].
Examples of applications that could use reliable bulk multicast
transfer include collaborative tools, distributed virtual reality,
and software upgrade services.
To meet the growing demand for reliable multicast, there is a large
number of protocol proposals. A few were published as RFCs before
the impact of congestion from reliable multicast was fully
appreciated, and these should be deprecated [DeprRFCs]. Two surveys
Mankin/Romanow/Bradner/Paxson [Page 2]
Internet-Draft Evaluating Reliable Multicast May 1998
of other publications are [DiotCrow97], [Obraczka98].
As we discuss in Section 3, the issues raised by reliable multicast
are considerably more complex than those related to reliable unicast.
In particular, in today's Internet, reliable multicast protocols
could do great damage through causing congestion disasters if they
are widely used and do not provide adequate congestion control.
Because of the complexity of the technical issues, and the abundance
of proposed solutions, we are putting in place review procedures that
are more explicit than usual. We compare this action with an IESG
action taken in 1991, RFC 1264 [Routing91], when community experience
with standard Internet dynamic routing protocols was still limited,
and extra review was deemed necessary to assure that the protocols
introduced would be effective, correct and robust.
Section 3 describes in detail the nature of the particular challenges
posed by reliable multicast. Section 4 describes the process for
considering reliable multicast solutions. Section 5 details the
additional requirements that need to be met by proposals to be
published as Standards Track RFCs.
3.0 Issues in Reliable Multicast
Two aspects of reliable multicast make standardization particularly
challenging. First, the meaning of reliability varies in the context
of different applications. Secondly, if special care is not taken,
reliable multicast protocols can cause a particular threat to the
operation of today's global Internet. These issues are discussed in
detail in this section.
3.1 One or Many Reliable Multicast Protocols or Frameworks?
Unlike reliable unicast, where a single transport protocol (TCP) is
currently used to meet the reliable delivery needs of a wide range of
applications, reliable multicast does not necessarily lend itself to
a single application interface or to a single underlying set of
mechanisms. For unicast transport, the requirements for reliable,
sequenced data delivery are fairly general. TCP, the primary
transport protocol for reliable unicast, is a mature protocol with
delivery semantics that suit a wide range of applications.
In contrast, different multicast applications have widely different
requirements for reliability. For example, some applications require
that message delivery obey a total ordering while others do not.
Some applications have many or all the members sending data while
others have only one data source. Some applications have replicated
Mankin/Romanow/Bradner/Paxson [Page 3]
Internet-Draft Evaluating Reliable Multicast May 1998
data, for example in an n-redundant file store, so that several
members are capable of transmitting a data item, while for others all
data originates at a single source. Some applications are restricted
to small fixed-membership multicast groups, while other applications
need to scale dynamically to thousands or tens of thousands of
members (or possibly more). Some applications have stringent delay
requirements, while others do not. Some applications such as file-
transfer are high-bandwidth, while other applications such as
interactive collaboration tools are more likely to be bursty but use
low bandwidth overall. These requirements each impact the design of
reliable multicast protocols in a different way.
In addition, even for a specific application where the application's
requirements for reliable multicast are well understood, there are
many open questions about the underlying mechanisms for providing
reliable multicast. A key question concerns the robustness of the
underlying reliable multicast mechanisms as the number of senders or
the membership of the multicast group grows.
One challenge to the IETF is to end up with the right match between
applications' requirements and reliable multicast mechanisms. While
there is general agreement that a single reliable multicast protocol
or framework is not likely to meet the needs of all Internet
applications, there is less understanding and agreement about the
exact relationship between application-specific requirements and more
generic underlying reliable mutlicast protocols or mechanisms. There
are also open questions about the appropriate integration between an
application and an underlying reliable multicast framework, and the
potential generality of a single applications interface for that
framework.
3.2 Congestion Control
A particular concern for the IETF is the impact of reliable multicast
traffic on other traffic in the Internet in times of congestion, in
particular the effect of reliable multicast traffic on competing TCP
traffic. The success of the Internet relies on the fact that best-
effort traffic responds to congestion on a link (currently as
indicated by packet drops) by reducing the load presented to the
network. Congestion collapse in today's Internet is prevented only
by the congestion control mechanisms in TCP, standardized by RFC 2001
[CongAvoid97, Jacobson88].
There are a number of reasons to be particularly attentive to the
congestion-related issues raised by reliable multicast proposals.
Multicast applications in general have the potential to do more
congestion-related damage to the Internet than do unicast
applications. One factor is that a single multicast flow can be
Mankin/Romanow/Bradner/Paxson [Page 4]
Internet-Draft Evaluating Reliable Multicast May 1998
distributed along a large, global multicast tree reaching throughout
the entire Internet.
Unreliable multicast applications such as audio and video are, at the
moment, usually accompanied by a person at the receiving end, and
people typically unsubscribe from a multicast group if congestion is
so heavy that the audio or video stream is unintelligible. Reliable
multicast applications such as group file transfer applications, on
the other hand, are likely to be between computers, with no humans in
attendance monitoring congestion levels.
In addition, reliable multicast applications do not necessarily have
the natural time limitations typical of current unreliable multicast
applications. For a file transfer application, for example, the data
transfer might continue until all of the data is transferred to all
of the intended receivers, resulting in a potentially-unlimited
duration for an individual flow. Reliable multicast applications
also have to contend with a potential explosion of complex patterns
of control traffic (e.g., ACKs, NACKs, status messages). The design
of congestion control mechanisms for reliable multicast for large
multicast groups is currently an area of active research.
The challenge to the IETF is to encourage research and
implementations of reliable multicast, and to enable the needs of
applications for reliable multicast to be met as expeditiously as
possible, while at the same time protecting the Internet from the
congestion disaster or collapse that could result from the widespread
use of applications with inappropriate reliable multicast mechanisms.
Because of the setbacks and costs that could result from the
widespread deployment of reliable multicast with inadequate
congestion control, the IETF must exercise care in the
standardization of a reliable multicast protocol that might see
widespread use.
The careful review and cautious acceptance procedures for proposals
submitted as Internet-Drafts reflects our concern to meet the
challenges described here.
4. IETF Process for Review and Publication of Reliable Multicast
Protocol Specifications
In the general case of individually submitted Internet-Drafts
(proposals not produced by an IETF WG), the process of publication as
some type of RFC is described in RFC 2026 (4.2.3) [InetStdProc96].
This specifies that if the submitted Internet-Draft is closely
related to work being done or expected to be done in the IETF, the
ADs may recommend that the document be brought within the IETF and
Mankin/Romanow/Bradner/Paxson [Page 5]
Internet-Draft Evaluating Reliable Multicast May 1998
progressed in the IETF context. Otherwise, the ADs may recommend
that the Internet-Draft be published as an Experimental or
Informational RFC, with or without an IESG annotation of its
relationship to the IETF context.
The procedure for Reliable Multicast proposal publication will have
as its default RFC status Experimental, when the technical criteria
listed in Section 5 are deemed to be fulfilled. Both the criteria and
the procedure reflect the AD's technical assessment of the current
state of reliable multicast technology. It does not reflect the
origins of the proposals, which we expect will be equally from
commercial vendors with initial products and from researchers.
Work on the development and engineering of protocols that may
eventually meet the review criteria could take place either in the
IRTF Reliable Multicast Research Group (http://www.irtf.org) or a
focused short IETF WG with an Experimental product.
When the work in reliable multicast technology has matured enough to
be considered for standardization within the IETF, the TSV Area may
charter appropriate working groups to develop standards track
documents. The criteria for evaluation of standards track technology
will be at least as stringent as those described herein (next
section).
5. Technical Criteria for Reliable Multicast
The Internet-Draft must (in itself or a companion draft):
a. Analyze the behavior of the protocol.
The vulnerabilities and performance problems must be shown through
analysis. Especially the protocol behavior must be explained in
detail with respect to scalability, congestion control, error
recovery, and robustness.
For example the following questions should be answered:
How scalable is the protocol to the number of users in a group,
number of groups, wide dispersion of group members? If
appropriate, how scalable is the protocol to the number of
senders?
Identify the mechanisms which limit scalability and estimate
those limits.
How does the protocol protect the Internet from congestion? How
well does it perform? When does it fail?
Mankin/Romanow/Bradner/Paxson [Page 6]
Internet-Draft Evaluating Reliable Multicast May 1998
Under what circumstances will the protocol fail to perform the
functions needed by the applications it serves?
Is there a congestion control mechanism? How well does it
perform? When does it fail? Note that congestion control
mechanisms that operate on the network more aggressively than
TCP will face a great burden of proof that they don't threaten
network stability.
b. Include a description of trials and/or simulations which support
the development of the protocol and the answers to the above
questions.
c. Include an analysis of whether the protocol has congestion
avoidance mechanisms strong enough to cope with deployment in the
Global Internet, and if not, clearly document the circumstances in
which congestion harm can occur. How are these circumstances to
be prevented?
d. Include a description of any mechanisms which contain the traffic
within limited network environments. It is likely that some
answers to a. and c. will mean that such mechanisms are required.
We recognize that the confinement of Internet applications is an
open research area.
e. Reliable multicast protocols must include an analysis of how they
address a number of security and privacy concerns. If the
protocol can be used in different modes of secure operation, then
each mode must be analyzed.
The analysis must document which of the various parties --
senders, routers (more generally, data forwarders), receivers,
retransmission sources -- must be trusted in order to ensure
secure operation and privacy of the transmitted data, to what
degree, and why. (One issue to address here are "man-in-the-
middle" attacks.)
To what degree can data be manipulated so that at least a
subset of the receivers receive different copies? Does the
protocol allow a group of receivers to determine whether they
all received the same data?
What limitations are placed on the retransmission mechanism to
prevent it from being abused to flood network links with
excessive traffic? Which parties must be trusted to ensure
this, and to what degree, and why? The presumption will be that
either a congestion control mechanism will inherently limit the
volume of retransmission traffic, and that this limiting
influence is robust under concerted attack; or that
Mankin/Romanow/Bradner/Paxson [Page 7]
Internet-Draft Evaluating Reliable Multicast May 1998
retransmission requests will be signed in a cryptographically
strong manner so that abuses of the mechanism can be traced
back to their source. Protocols that do not provide either of
these forms of protection face a great burden of proof that
they don't threaten network stability.
What sort of key management does the protocol require, and
provide for?
6. Security Considerations
This memo specifies in Section 5.e. that reliable multicast Internet-
Drafts reviewed by the Transport Area Directors must explicitly
explore the security aspects of the proposed design.
7. Acknowledgments
Sally Floyd, Steve McCanne, Mark Handley, Steve Bellovin and Mike
Reiter gave especially helpful comments on drafts of this document.
8. References
[RMMinutes 1997], Minutes the Second Reliable Multicast Research
Group Meeting. September 1997. http://www.east.isi.edu/rm
[Floyd97] Floyd, S., Jacobson, V., Liu, C., McCanne, S., and Zhang,
L., A Reliable Multicast Framework for Light-weight Sessions and
Application Level Framing. IEEE/ACM Transactions on Networking,
December 1997 An online version of the paper is at
http://ee.lbl.gov/floyd/srm-paper.html.
[InetStdProc96] Bradner, S., The Internet Standards Process --
Revision 3, RFC 2026, October 1996.
[DiffServBOF97] [6] http://www.ietf.org/proceedings/97apr -
Transport Area - FDDIFS BOF, April 1997.
[DeprRFCs]. Freier, A. Multicast Transport Protocol, RFC 1301,
February 1992. and Braudes, R., Zabele, S., Requirements for
Multicast Protocols, RFC 1458, May 1993.
[DiotCrow97], Diot, C., Crowcroft, J., Multicast Transport Survey.
Journal of Selected Areas in Communications, 1997.
[Obraczka98] Obraczka, K., Multicast Transport Mechanisms: A Survey
and Taxonomy. To appear in IEEE Communications, 1998.
Mankin/Romanow/Bradner/Paxson [Page 8]
Internet-Draft Evaluating Reliable Multicast May 1998
[Routing91] Hinden, R. M., Internet Engineering Task Force internet
routing protocol standardization criteria, RFC 1264, October 1991
[CongAvoid97] Stevens, W. R., TCP Slow Start, Congestion Avoidance,
Fast Retransmit, and Fast Recovery Algorithms, RFC 2001, January 1997
[Jacobson 1988] Jacobson, V., Congestion Avoidance and Control,
Proceedings of SIGCOMM '88, August 1988, pp. 314-329. An updated
version of this paper is available at
"ftp://ftp.ee.lbl.gov/papers/congavoid.ps.Z".
9. Authors Addresses
Allison Mankin - Past TSV Area Director
USC/ISI East
4350 N. Fairfax Dr., Suite 620
Arlington VA 22203
USA
703 812 3706
mankin@east.isi.edu
Allyn Romanow - Past TSV Area Director
MCI Corporation
2560 North First Street
San Jose, CA 9531
USA
408 922 7143
allyn@mci.net
Scott Bradner - TSV Co-Area Director
Harvard University
1350 Mass. Ave., Rm 876
Cambridge MA 02138
USA
617 495 3864
sob@harvard.edu
Vern Paxson - TSV Co-Area Director
MS 50B/2239
Lawrence Berkeley National Laboratory
University of California
Berkeley, CA 94720
USA
510-486-7504
vern@ee.lbl.gov
Mankin/Romanow/Bradner/Paxson [Page 9]
Internet-Draft Evaluating Reliable Multicast May 1998
Mankin/Romanow/Bradner/Paxson [Page 10]