March 25 to March 27, 1998 the Internet Architecture Board (IAB) held
a workshop on Routing. The workshop focused on current problems
within the Internet and the long term solutions that should be
addressed. This document summarizes the discussions the group had on
routing, and lists the conclusions reached by the workshop. Section
2 lists the conclusions reached by the participants of the workshop
and the suggestions for additional work or redirection of current
work. Sections 2.1-2.10 attempt to extract the major points of what
was, in actuality, many multifaceted discussions, sometimes occurring
all at the same time. Appendix A contains a list of the participants
who attended the workshop. The full body of the report can be found
at http://www.iab.org.
The topics covered at length during the IAB workshop were:
1. Scaling of Unicast Routing and Addressing (section 2.1)
2. Unicast Addressing Issues (Section 2.2)
3. The Effect of extending IP version 4 in the Internet by using
Network Address Transformation boxes (Section 2.3)
4. Multicast Routing (Section 2.4)
Deering, et al. Informational [Page 2]
RFC 2902 Overview of the 1998 IAB Routing Workshop August 2000
5. Routing Instability (Section 2.5)
6. Quality of Service Routing (Section 2.6)
7. Routing Security (Section 2.7)
8. BGP Policy (Section 2.8)
9. Flows of information from network routing to hosts for improved
services (Section 2.9)
In addition the following topics were briefly covered:
a. Multi-strand trunking
b. Better tools for monitoring and diagnosis of network problems
c. Routing protocol bandwidth minimization
d. Automatic renumbering and automatic organization
e. Anycast
f. Load-sensitive routing
g. Geographical addressing
These shorter topics are contained in section 2.10.
It would be unrealistic to assume that the workshop had definitive
answers to all the technical problems that were raised. The best
that can be hoped is that we raised most of the relevant issues and
gave opinions that were the best guess of the people at the meeting,
keeping in mind that the attendees did not come armed with data to
back up opinions. Much of the discussion amounted to an exploration
of the intuition of the experts in attendance, intuition gained after
years of experience in making the Internet work. More work is needed
to validate the intuition and experience by way of scientific
experimentation and analysis. Unfortunately, it's not so easy to
find a spare collection of global Internets upon which one might
perform controlled experiments.
The participants came to a number of conclusions after the
discussions referred to in sections 2.1-2.10. These conclusions,
presented in this document, provide summary statements and action
items for the IETF community.
The participants of the workshop came to the following conclusions
1. Most of the current unicast routing stability problems can be
fixed with improved implementation.
Deering, et al. Informational [Page 3]
RFC 2902 Overview of the 1998 IAB Routing Workshop August 2000
2. Some long term systemic issues that may eventually overwhelm the
unicast routing are:
- Flaps - which will only get worse unless work is undertaken
- Multi-homing
3. We'd like more research into what's breaking; not just more data,
but more analysis of the data
The group reviewed the following potential solutions:
- Architected NAT (improving the existing Network Address
Translation schemes to provide better scaling)
- IPv6 (deploying an IP version 6 infrastructure)
- MAP/Encap (map to aggregatable addresses and encapsulate the
original packet)
- Do nothing
- Aggressive renumbering (try to continue to encourage renumbering
to improve utilization of the IP version 4 address space)
- Metro addressing (use a geographical or metropolitan based
addressing scheme)
Levels of hierarchy do not matter to the customers. Address
hierarchy must be distinguished from routing hierarchy. The group
examined whether the current Internet has enough levels of hierarchy
in Internet addresses or routing infrastructure. The group did not
find that levels of hierarchy should be added to the Internet, at
least for now. Flat routing at the AS level seems to be workable; if
this changes in the future, hierarchy would need to be revisited, and
studied with due consideration to convergence time for routing
algorithms and trust management. There is no universal agreement
that adding levels of hierarchy at this point in time provides a
well-defined benefit. Furthermore, two levels is difficult for many
people, and any more than that is difficult both to build and to use.
Deering, et al. Informational [Page 4]
RFC 2902 Overview of the 1998 IAB Routing Workshop August 2000
Upon reviewing the NATs, the group
1. Noted that NAT devices are fairly widely deployed
2. Identified various problems with the use of NAT devices within
the internet
3. Discussed the interaction between NAT devices and applications
4. Listed the following options regarding NAT devices:
- Eliminate NATs
- Fix NATs to interact better with the rest of the Internet
- Fix applications to interact better with NAT boxes
- Don't do certain things -- like IP Security (IPSec)
1. Forward our concerns, problems and suggestions to the appropriate
working groups
2. Note architectural work outside the NAT working group
3. Suggest to the IAB that it continue to be concerned about the
issues involving NATs
Since the multicast model was created, many multicast applications
have been tried over the Internet multicast routing fabric. The
group began to discuss the multicast model in terms of enabling
multicast applications to run efficiently, and scale favorably with
future growth. Multicast applications place varying requirements on
multicast routing.
Multicast applications may have a variable:
- number of sources,
- number of receivers,
- amount of data,
- amount of data in a burst, and length of quiet periods
- number of groups utilized per application or per set of
cooperating applications, and
- amount of time during which the group exists
- topological distance between members of the group.
- volatility of membership
Deering, et al. Informational [Page 5]
RFC 2902 Overview of the 1998 IAB Routing Workshop August 2000
Multicast routing must provide the flexibility to support the varying
requirements of different multicast applications. The current
multicast model establishes multicast routing paths upon reception of
a data packet. The discussion on the viability of the multicast
model examined the viability of the model in terms of the uses of
multicast routing by applications and the scalability to full
Internet usage. For example, providing for many groups of small
conferences (a small number of widely-dispersed people) with global
topological scope scales badly given the current multicast model.
The group felt the existing multicast protocols and multicast should
be evaluated in terms of the requirements listed above. The group
suggested that the evaluation should include the multicast protocols
DVMRP [12], MOSPF [8], PIM [4], CBT [2], and Express [5], as well as
the following mechanisms used by multicast applications:
1. Registering with the core or the RP (Rendezvous Point),
2. Having the ID of the group include the core, and having joins
specify the core
3. Having the ID of the group include the core, and having joins
and data specify both
4. Sending data via unicast to all members, and
5. Sending data via unicast transport to the RP.
The group acknowledged that the current multicast model does not
scale well for all scenarios that applications use.
The group noted that reliable multicast is surprisingly orthogonal to
the issues about the scaling of the multicast model to all possible
applications.
Encourage evaluation and written reports on these multicast
protocols, and mechanisms for different types of protocols.
Notify the IRTF Routing Research Group of the need to charter
activity in this area.
Damping the effects of route updates enhances stability, but possibly
at the cost of reachability for some prefixes. A prefix can be
damped and reachable via another path, so that for such prefixes the
effects of damping are less serious than for other prefixes. The
performance of various algorithms for enhancing stability should be
Deering, et al. Informational [Page 6]
RFC 2902 Overview of the 1998 IAB Routing Workshop August 2000
measured by recording whether the affected route prefixes are
reachable or not reachable. Using current damping approaches,
approximately 1% of the prefixes are affected at any one point in
time. We should try to find out how many prefixes are unreachable
because of damping.
The conclusion is that this effort merits continued investigation.
The IRTF Routing Research Group should measure how stable things are,
and if stability is an issue, to study methods of making them more
stable.
The group noted that the terms Type of Service (ToS), Class of
Service (CoS), and Quality of Service (QoS) are imprecise as
currently used. The discussion started by defining the terminology
as follows:
ToS: hop by hop routing based on destination plus ToS bits [9]
CoS: classes of service based on service contracts. These classes
of service are enabled by a variety of mechanisms which include
queueing, and multiple physical or link level paths.
QoS: managing routes that meet certain quality of service constraints,
and involving the following steps:
* routing the resource requests
* setting up a path that satisfies the constraints
* routing the data
There is no smooth dividing line between between ToS and QoS. ToS is
relative. QoS is absolute. The group discussed whether there is a
demand for ToS, CoS and QoS. Differentiated-services [3] as discussed
in the IETF is ToS++.
The group also discussed a more general concept of "Constraint Based
Routing" which was defined as traffic engineering on large aggregated
flows. Constraint based routing allows the providers to better
utilize the bandwidth in their network to handle traffic requests
from users. Besides enabling policy management techniques,
constraint based routing allows providers to route traffic based on
the characteristics of the traffic flows.
Deering, et al. Informational [Page 7]
RFC 2902 Overview of the 1998 IAB Routing Workshop August 2000
After a lengthy discussion of the various problems of network
security, the group notes that:
1. Routers need intrinsic system security as good as or better than
any host computer.
2. Improving router security will not solve all problems.
3. Console access to the router can do everything.
4. One compromised router can create disaster.
5. ISPs and vendors should consider taking some control traffic out
of band, due to lack of wire speed authentication.
6. We discussed other issues that will be passed on to the
appropriate people involved with network security.
7. Identified areas of work to improve things (e.g., wire speed
authentication).
The IETF should encourage work on "wire speed" authentication, pair-
wise authentication of routers in routing protocols, and Byzantine
robustness [6] in routing protocols.
During our discussion on routing policy the group reviewed what could
be done with BGP. The group noted that:
1. Some routing policies requested by ISPs or NSPs are not solvable
with BGP. Some of these "unsolvable" routing policies can be put
into effect using tunnels and static configuration.
2. BGP is only a mechanism for announcing reachability
3. BGP routing controls traffic direction without regard to traffic
volume.
4. BGP policy management is too delicate, too easy to mess up, and
fragile.
Deering, et al. Informational [Page 8]
RFC 2902 Overview of the 1998 IAB Routing Workshop August 2000
5. Router Configuration Language is very complex and error-prone
6. We can't count on symmetric routing, so ISPs/NSPs/Enterprise nets
should deal with it.
The group concluded the Internet needed a better routing policy
specification language.
Publishing information about traffic statistics along backbone routes
could improve the way Internet services replicate data for retrieval
from various sites. This replication could be especially important
for the retrieval of information off the web. Currently, web pages
refer people to caches local to their sites; for instance, a European
site might be used for United Kingdom customers and a North American
site for North American customers. Proponents of web caches want to
auto-configure the locations of web caches so a user's web browser
can automatically discover the local cache. Other applications share
this need for finding the best cache for a particular service.
The group recommends a BOF be held on Measuring Path Characteristics.
Measurement of path characteristics should include:
- format for exchange of measurement data
- mechanisms for distribution of measurement data
IPPM working group [7] is dealing with issues within the measurement
problem space.
PPP did multi-link in a way that required too much computation and
could not be used for faster links. Internet technology should treat
multiple parallel trunks as 1 link at the IP layer, but with multi-
dimensional metrics.
Deering, et al. Informational [Page 9]
RFC 2902 Overview of the 1998 IAB Routing Workshop August 2000
Multi-strand Trunking - Action Items
There is design and development work at layer two which should be
done to support the multiple parallel trunks. This layer two work
is outside the scope of the IETF. Layer three routing should
support richer metrics in OSPF.
1. It would be nice to have an Authoritative Database listing those
prefixes permitted from each AS. The authoritative data base was
attempted before without success, but the group felt it might be
useful to try again.
2. SNMP version 3 should be deployed in order to make use of its
improved authentication, scope and rate limiting
3. Remotely-controlled traffic monitors should be used to measure
traffic
4. Better tools are needed for preventative problem detection
1. Encouraged an authoritative database within the Internet
2. Notify SNMP version 3 working groups regarding needs for
authentication, scope, and rate limiting.
3. Encourage funding of better tools for remotely controlled traffic
sources and pro-active problem detection.
A BOF should be held to plan work on anycast.
If a working group forms, a paper on the advantages and disadvantages
of anycast should be included as part of the charter.
Deering, et al. Informational [Page 10]
RFC 2902 Overview of the 1998 IAB Routing Workshop August 2000
While load sensitive routing is interesting in some ways, it cannot
be considered until certain problems are worked out. Currently,
constraint based routing is assigning administrative metrics to allow
routing to adapt to different traffic patterns. Load sensitive
routing may increase oscillation and instability of routes. This
instability of routes, sometimes called churn, may affect the ability
of the routing infrastructure to scale.
Load sensitive routing would allow IGPs to better utilize links.
Past and current efforts in load sensitive routing include: QoS OSPF
[10], Q-OSPF [10], and load sensitive routers developed by BBN.
The IRTF Routing Research group chair and Routing Area Director
should discuss this subject and determine what techniques from Load
Sensitive IGP routing are ready for IETF, and what requires
additional research.
1. The IAB should be concerned about the issues involving NATs
2. Authoritative Database (for addresses within domains) should be
encouraged within the Internet
3. Encourage funding of better tools for remotely controlled traffic
sources and pro-active problem detection.
1. NAT: Forward our concerns, problems and suggestions to the
appropriate working groups
2. We recommend that IETF should work the issue of Constraint Based
Routing.
3. The IETF should encourage work on "wire speed" authentication,
pair-wise authentication of routers in routing protocols, and
Byzantine robustness in routing protocols.
Deering, et al. Informational [Page 11]
RFC 2902 Overview of the 1998 IAB Routing Workshop August 2000
4. Concerns about the Routing Policy Specification Language (RPSL)
should go to the Routing Policy Systems (RPS) working group chair.
5. The group recommends a BOF be held on Measuring Path
Characteristics. The BOF should consider the data exchange format
of measurement and mechanisms to distribution of data mechanism.
It is noted that the IPPM working group is dealing with issues
within the measurement problem space.
6. There is layer two work which should be done to support the
multiple parallel trunks which is outside the scope of the IETF.
Layer three routing should support richer metrics in OSPF.
7. SNMP version 3 working groups should be notified about the issues
about authentication, scope, and rate limiting.
8. A BOF should be held to plan work on anycast. A document on
anycast should be part of the proposed working group charter.
1. We recommend that the IRTF Routing Research working group try to
encourage more analysis of routing data, not just the collection
of more data.
2. Encourage evaluation and written reports on the evaluation of
multicast protocols and mechanisms for different types of
protocols
3. The IRTF Routing Research group chair and the Routing Area
Director should discuss Load Sensitive IGP routing and determine
whether it is ready for the IETF.
Security considerations were an important part of the discussions at
the workshop, but the workshop decided not to publish a summary of
these discussions. Other documents that address the issues of
routing infrastructure security have recently been published.
(Email addresses as of the meeting date.)
Harald Alvestrand Harald.Alvestrand@maxware.no
Fred Baker fred@cisco.com
Jeff Burgan burgan@corp.home.net
Brian Carpenter brian@hursley.ibm.com
Noel Chiappa jnc@ginger.lcs.mit.edu
Rob Coltun rcoltun@fore.com
Steve Deering deering@cisco.com
Deborah Estrin estrin@usc.edu
Deering, et al. Informational [Page 12]
RFC 2902 Overview of the 1998 IAB Routing Workshop August 2000
Dino Farinacci dino@cisco.com
Paul Francis francis@slab.ntt.co.jp
Elise Gerich epg@home.net
Joel Halpern jhalpern@newbridge.com
Sue Hares skh@merit.edu
Cyndi Jung cmj@3Com.com
Dave Katz dkatz@jnx.com
Tony Li tli@juniper.net
Peter Lothberg roll@stupi.se
Louis Mamakos louie@uu.net
Dave Meyer dmm@cisco.com
Keith Moore moore@cs.utk.edu
Bob Moskowitz rgm@htt-consult.com
Thomas Narten narten@raleigh.ibm.com
Vern Paxson vern@ee.lbl.gov
Charles E. Perkins cperkins@eng.sun.com
Radia Perlman Radia.Perlman@East.Sun.COM
Yakov Rekhter yakov@cisco.com
Allyn Romanow allyn@MCI.NET
Martha Steenstrup msteenst@bbn.com
George Swallow swallow@cisco.com
References
[1] Alaettinoglu, C., Bates, T., Gerich, E., Karrenberg, D., Meyer,
D., Terpstra, M. and C. Villamizar, "Routing Policy
Specification Language (RPSL)", RFC 2280, January 1998.
[2] Ballardie, A., "Core Based Trees (CBT) Multicast Routing
Architecture", RFC 2201, September 1997.
[3] Blake, S., Black, D., Carlson, M., Davies, E., Wang, Z. and W.
Weiss, "An Architecture for Differentiated Service", RFC 2475,
December 1998.
[4] Estrin, D., Farinacci, D., Helmy, A., Thaler, D., Deering, S.,
Handley, M., Jacobson, V., Liu, C., Sharma, P. and L. Wei,
"Protocol Independent Multicast-Sparse Mode (PIM-SM): Protocol
Specification", RFC 2362, June 1998.
[5] Holbrook, H., Cheriton, D, "EXPRESS Multicast", SIGCOMM 99,
September 1999.
[6] Charlie Kaufman, Radia Perlman, and Mike Speciner. Network
Security: Private Communication in a Public World, pages 462--
465. Prentice-Hall, Inc., 1995.
Deering, et al. Informational [Page 13]
RFC 2902 Overview of the 1998 IAB Routing Workshop August 2000
[7] W. Leland and M. Zekauskas (chairs). IP Performance Metrics
(IPPM), October 1997. http://www.ietf.org/html.charters/ippm-
charter.html.
[8] Moy, J., "Multicast Extensions to OSPF", RFC 1584, March 1994.
[9] Nichols, K., Blake, S., Baker, F. and D. Black, "Definition of
the Differentiated Services Field (DS Field) in the IPv4 and
IPv6 Headers", RFC 2474, December 1998.
[10] H. Sandick and E. Crawley (chairs). QoS Routing (qosr), April
1997. http://www.ietf.org/html.charters/qosr-charter.html.
[11] C. Villamizar and C. Alaettinoglu (chairs). Routing Policy
Syntax (RPS), July 1995. http://www.ietf.org/html.charters/rps-
charter.html.
[12] Waitzman, D., Partridge, C. and S. Deering, "Distance Vector
Multicast Routing Protocol", RFC 1075, November 1988.
Deering, et al. Informational [Page 14]
RFC 2902 Overview of the 1998 IAB Routing Workshop August 2000
Authors' Addresses
Questions about this memo can be directed to:
Stephen E. Deering
Cisco Systems, Inc.
170 West Tasman Drive
San Jose, CA 95134-1706
USA
Phone: +1 408 527-8213
EMail: deering@cisco.com
Susan Hares
Merit, Inc.
1071 Beal Avenue,
Ann Arbor, MI 48109
USA
Phone: +1 313 936-2095
EMail: skh@nexthop.com
Radia Perlman
Sun Microsystems Laboratories
2 Elizabeth Drive
Chelmsford, MA 01824
USA
Phone: +1 978 442-3252
EMail: Radia.Perlman@sun.com
Charles E. Perkins
Nokia Research Center
313 Fairchild Drive
Mountain View, CA 94043
USA
Phone: +1 650 625-2986
EMail: Charles.Perkins@nokia.com
Fax: +1 650-625-2502
Deering, et al. Informational [Page 15]
RFC 2902 Overview of the 1998 IAB Routing Workshop August 2000
Full Copyright Statement
Copyright (C) The Internet Society (2000). All Rights Reserved.
This document and translations of it may be copied and furnished to
others, and derivative works that comment on or otherwise explain it
or assist in its implementation may be prepared, copied, published
and distributed, in whole or in part, without restriction of any
kind, provided that the above copyright notice and this paragraph are
included on all such copies and derivative works. However, this
document itself may not be modified in any way, such as by removing
the copyright notice or references to the Internet Society or other
Internet organizations, except as needed for the purpose of
developing Internet standards in which case the procedures for
copyrights defined in the Internet Standards process must be
followed, or as required to translate it into languages other than
English.
The limited permissions granted above are perpetual and will not be
revoked by the Internet Society or its successors or assigns.
This document and the information contained herein is provided on an
"AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING
TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING
BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION
HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF
MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
Acknowledgement
Funding for the RFC Editor function is currently provided by the
Internet Society.
Deering, et al. Informational [Page 16]