IPv6%20Benchmarking%20Methodology

1
IPv6 Benchmarking Methodology

• Ciprian Popoviciu, Ahmed Hamza, Gunter Van de
  Velde, Diego Dugatkin

IETF 68, March 20th 2007 Prague
2

• Agenda
• Overview
• Last Call comments
• Next steps

3
A document providing guidance in the area of
IPv6 benchmarking would be welcome to
organizations (including the US federal agencies
mandated to deploy IPv6 on their backbone
networks) attempting to understand why and how
network device IPv6 performance must be tested. A
document that attempts to define which areas need
to considered and which describes how to
test/benchmark these areas may be well received.
Bill Cerveny

• Overview
• Complements RFC2544 by
• Adds benchmarking methodology recommendations
  that address specific aspects of IPv6 protocol
  architecture
• Provide an updated list of benchmarks based on
  the experience gained with applying the RFC2544
  recommendations to IPv4
• Adds information related to SONET as a popular
  media type not mentioned by RFC2544
• Went through several rounds of review within BMWG
  and v6ops WG

4
Overview (cont.)

• Interest
• Expressed by networking and test tool vendors
• There is at least one implementation of the draft
  in a benchmarking suite
• Used in benchmarking related to the OMB and DoD
  IPv6 mandates.
• Timeline
• Voted WG Working Item during the Montreal IETF
  66th meeting
• Currently in Last Call until March 20

5
Actions on Last Call Comments

• The LC comments relate primarily to frame size
  considerations (which are IP version agnostic)
  and not to the IPv6 specific benchmarking
  recommendations.
• Based on the comment from David Newman and in
  observance of RFC1242, we will change "maximum
  throughput" to "throughput" in Appendix A.1 and
  A.2.
• Based on David Newman's suggestion, we will round
  down the theoretical rates listed in Appendix A.1
  and will include a note regarding this round down
  as well as a reference to the tolerance due to
  clock slip

6
Actions on Last Call Comments (cont.)

• Minimum Frame size for SONET. One proposal is 48
  bytes. Based on RFC2544, IPv6 main header 40
  bytes 8 bytes UDP header (ECHO) for 56 bytes.
• Maximum frame size for SONET. In the Appendix we
  list 4096 bytes. Proposal is to add the note
  Larger values might be supported by the various
  products. The maximum supported frame size should
  be used in benchmarking in addition to the
  recommended frame sizes."
• Maximum frame size for Ethernet. The Appendix
  lists frame sizes up to 9216 bytes. Should we
  include 2000 explicitly in order to have the data
  point for the size recommended by 802.3as?

7
Next Steps

• The LC comments will be integrated in version -02
• Close LC

8

• draft-ietf-bmwg-ipv6-meth-01
• THANK YOU!

9

• Document Goals
• Address a very acute and growing need for
  recommendations on evaluating network element
  performance for IPv6 deployments
• A complement rather than a replacement of RFC2544
  in accordance with BMWG strategy
• Provide the additional, IPv6 specific guidelines
  to IP benchmarking while indicating the aspects
  of RFC2544 that are IP version independent
• Maintain the structure and spirit of RFC2544.

10
WG Feedback

• List of Reviewers from BMWG and V6OPS WGs Scott
  Bradner, Al Morton, Fred Baker, Pekka Savola,
  Brian Carpenter, Tim Chown, Benoit Lourdelet,
  Daniel Roesen, Jerry Perser, William Cerveny ,
  Athanassios Liakopoulos, Rajiv Papneja, Sven
  Lanckmans, Silvija Dry, Aamer Akhter, Rajiv
  Asati, David Newman, Jim Mcquaid, Timmons Player,
  Miles McCredie, Curtis Villamizar.
• IPv6 and Test tools experts reviewed and provided
  valuable feedback off the BMWG alias.
• Thank you to all reviewers!