CURRENT MEETING REPORT

Minutes of the Realtime Traffic Flow Measurement Working Group (rtfm)

Reported by Cynthia Mills, GTE Laboratories, Inc. and Nevil Brownlee, University of Auckland

The group's charter and milestones were reviewed and confirmed. Nevil Brownlee gave a brief overview of the Traffic Flow Measurement model, and of NeTraMet (his public-domain implementation). Sig Handelman gave a status report of the IBM implementation.

Existing work on traffic flow measurement was discussed, particularly:

• The work of the NLANR group at the San Diego Supercomputer Centre (see http://www.nlanr.net/NA for a good introduction).
• A journal article on "Measurement, Modelling and Emulation of Internet Round Trip Delays" was discussed. The reference has been posted to the mailing list.
• The Flow Statistics provided by Cisco's Flow switching option for their high-end routers.
• The work of other IETF Working Groups such as BMWG/IPPM, RMON and RSVP.

The distinction between this group and IPPM was discussed. IPPM is concerned with measurements which a user may use EXTERNALLY to measure an ISP's performance, while RTFM provides instrumentation which may be used INTERNALLY in an ISP's network to measure flows and performance. Steve Corbato gave a short presentation of his recent work on high-speed polling of router variables.

We will publish a detailed list of references to these on the RTFM Web page (http://www.auckland.ac.nz/net/Internet/rtfm/TOP.html).

The 'Flow Measurement: Architecture' Draft was reviewed. Many people asked interesting questions, but only editorial changes were requested. We will run a two-week call on this draft, make these changes so as to produce a new Draft early in March. After a two-week last call we will submit this Draft to IESG for publication as an experimental RFC.

The 'Flow Measurement: Meter MIB' document was reviewed. A number of significant changes to this have been suggested since it was published in mid-February. These will be made, and a new draft published. If - after a two-week last call - there are no requests for further changes, we will submit this Draft to IESG for publication as an experimental RFC.

Two further new Drafts, 'Flow Measurement: Background' and 'Flow Measurement: Experience' will be prepared before the Montreal IETF meeting. These are intended for publication as information RFCs.

The group's next work item, a revised version of the traffic flow model, was considered. Two topics of interest were raised:

• How well does the meter cope with running out of resources? Switching to a standby rule set so as to reduce the rate new flow records are created has worked well in practice, but we should consider sending an alarm request to the manager.

• Would it be sensible to have a hard disk in the meter? This would make the meter less simple than it is now, but it would let the meter save its configuration data such a rule sets so that it could restart after a power outage without needing a download from the manager. It would also allow the meter to write its flow table to disk when a manager commanded it to do so.

We will collect further ideas on the mailing list for discussion in Montreal.