2.7.16 Deployment Considerations of Implementing Differences (decides) bof

Current Meeting Report

Held July 14, 15.30- 17.45

Deployment Considerations of Implementing Differentiated Services

About 200 attendees, few left til the very end, which showed a very high level of interest since the room was unbearably hot and crowded.

slides are (or will be) at http://www.cs.ucl.ac.uk/research/decides/

Kathie Nichols - 15 mins

Overview -
- Victor Firoiu
- Sylvie Giordano, EPFL

"Effect of Number of Drop Precedences in Assured Forwarding"
- Raj Jain

"Study of TCP and UDP Interaction for the AF PHB"
- Nabil Seddigh

Linux work
- Jamal Hadi Salim - 15 mins
- IBM Edge router experience
- Steve Nadas nadas@raleigh.ibm.com - 5 mins

Applicability Discussion - All 30 mins
- Roland Bless & Klaus Wehrle - 15 mins

Noriko Hata (DISCO proposer) - didn't show

signaling/adission with 2 bits:
- Borje Ohlman <Borje.Ohlman@abc.se>, turanyi@ttt-atm.ttt.bme.hu,
- Lars.Westberg@era-t.ericsson.se

Wrapup and Output Decisions - Jon Crowcroft 10 mins

We expected to see three approaches, simulation, measurement and analysis.

For single node implementation case, we saw some reports and measurement results, though for full implementation across a network there wasn't anything reported. It was felt that its too early, but that around 1 year from now, we might expect to see a requirement for a dboned working group (by analogy with mboned), if there is a requirement for IETF input on standards for tools to do measurement and debugging of deployed implementations.

We had a strong goal which was to set a realistic goal for useful input from simulations, and its my feeling that we conveyed this well; clearly all the early work from simulation tends to be on simple traffic sources and toplologies. We hope that if the BOF repeats in DC we will see more mature results in this area. If folks have sufficient generality i ntheir results, it was expected that they might contribute to the community via an informational RFC, but that a WG was not needed for this.

Analytic work appears to be making progress. There was a good consenus amongst those interested that an IRTF Research Group could help focus the input from that community (or even help form that community) towards IETF problems.

The consensus of the people in the room was that the BOF should meet again in DC if permitted.


Applicability Considerations of Differentiated Services
IBM DiffServ Implementation Experience on Edge Routers
Network Topologies for Evaluating Diffserv