BMWG Minutes The meeting was chaired by Kevin Dubray. Contributors to the meeting's minutes include Scott Poretsky, Al Morton, and Kevin Dubray. About 35 people attended the BMWG session. The original agenda, 0. Administration 1. Benchmarking Network-layer Traffic Control Mechanisms Attempt to resolve contentious issues regarding the definition of the term "congestion." Assess I-D's last call suitability. 2. Status update of OSPF benchmarking I-Ds: 3. Followup to a proposed, new BMWG work item: "SONET/SDH APS Performance Benchmarking." was approved, with the modification that the chair would attempt to briefly address item 2, the OSPF benchmark I-Ds, as the I-D's principals were unable to attend this session. 0. Administration ----------------- A status summary of BMWG Internet Drafts was presented as follows: *AD/IESG Review: , in review , editing required *I-D Last Call: , ready to forward to ADs. , revised *I-Ds: , revised , revised , revised initial *Other , expired Kevin Dubray made a note that we'd try to get the FIB methodology draft 'unexpired' and in last call, as there was not a lot of commentary on it. Regarding the BGP convergence terminology work, it was noted that the AD/IESG review had presented substantive commentary. It was unclear as to where the principals were in addressing the draft. Elwyn Davies replied that it was on the radar and he would help coax the production of a follow-up I-D. There was concern raised over the progression (i.e., slow progress of the multicast methodology I-D. Dubray deferred the discussion of this topic until later in the meeting. There was a re-ordering of the agenda slightly to talk about the OSPF benchmarking work, before tackling the subsequent meeting items. Status update of OSPF benchmarking I-Ds. --------------------------------------------------- Due to other WG meeting conflicts, the OSPF I-D editors were unable to attend the BMWG session. Dubray said that the editors contacted him prior to the meeting and indicated that the terminology and methodology were nearing readiness for WG last call. There appeared to be agreement in the room along those lines. Kevin also asked folks to pay particular attention to the corresponding "applicability" statement, as it represented a new type of document for the BMWG - trying to identify contexts where the benchmarks had utility and where they did not. Benchmarking Network-layer Traffic Control Mechanisms ----------------------------------------------------- Jerry Perser led a discussion on the current state of the terminology I-D. The slides can be found in the proceedings. A discussion on the I-D's definition of "Congestion" ensued. Jerry presented a viewpoint on the nature of delay and congestion. In the I-D, Congestion was tightly coupled to lossful events. It was argued that congestion can mean more than this. If the I-D wanted to stick to this definition, it should not have a title so encompassing. It was recommended that definition be called something like "Loss-indicated Congestion" to prevent confusion. With respect to delay, it was discussed that delay may also be an indicator of congestion, but it was not necessarily a deterministic one. So, it was recommended to add a paragraph to the Congestion discussion on why delay is not a deterministic indicator of congestion and add a paragraph to the Delay discussion on what's different from RFC 2679. In delay calculations, Jerry made note that the inclusion of packet insertion time generally concerned him. It was also thought to be useful to add a paragraph comparing the notion of a "Loss Vector" to RFC 2680. Proposed BMWG work item: "SONET/SDH APS Performance Benchmarking." ------------------------------------------------------------------- Takumi Kimura and Jerry Perser were present to lead a discussion regarding the I-D. Presenters' slides can be found in the proceedings. Following the general presentation, a discussion was had regarding whether this work is a good fit for the BMWG. A question was asked as to whether this work was better suited for sub-IP types of organizations, like the IEEE. It was countered that providers are asking vendors to quantify how sub-IP technologies' failover or recovery affect the IP layer related reachability and forwarding. It was thought the work was OK, if it kept its focus on IP affected responses. It was thought the title should be changed to reflect said focus. Since protection could happen at various layers (or sub-layers), such as SONET or MPLS, there was a discussion on how best to tackle this. The group gravitated to the idea that a single, common terminology document with subsequent methodology documents for individual recovery mechanisms (e.g., one for MPLS, one for APS, etc.) might be the way to go. A specific comment directed towards the draft thought it would be a good thing to lose the use of the term "service". I.e., reformulate the notion of "Loss of Service," or "Loss of Service Time." Proposed BMWG work item: Core Router Accelerated Life Testing ------------------------------------------------------------- Scott Poretsky gave an overview of the proposed work item. His slides are in the proceedings. Many ISP folks in the room indicated they do forms of this style of testing. It was also thought that while the tests addressed in the I-D were interesting, there was a question of "fit" for the work in the BMWG. Some issues raised: The tests seem to border more on functional tests than benchmarks. (The IETF has seemed to shy away from specifying functional assessments.) The tests seem to rely on white box instrumentation. (The BMWG has traditionally agreed that white box tests are out of scope for the WG.) The bounding of the test domain. What do you include? Where do you stop? The difficulty "standardizing" an operational model on which to test. Parallels were drawn to the failed Router Benchmarking Framework effort. While there appeared to be general interest in the work from the attendees, the chair counsel that work might have to be more tailored along traditional BMWG lines to be considered. With the discussion on the Router Life Testing concluded, discussion returned to the Multicast Benchmarking Methodology I-D. There was some frustration expressed as to the degree of progress the draft was making. The chair replied that, currently, the only topic being actively discussed on the mailing list was the encapsulation family of metrics. It was countered that the methodology I-D seemed disconnected from its terminology parent, RFC 2432. Kevin made the appeal to get issues and suggested fixes to the list so we can bring this work item to a quick and useful conclusion.