AGENDA IPPM Meeting @ IETF81 ============================ Monday, 9:00 - 11:30 0. 5' Administrativia Scribes: Dave McDysan, Al Morton (Minutes, Scribe, Blue Sheet, Note Well, ...) - See Slides. 1. 5' Status of drafts not discussed today - Round Trip Loss stauts (Al Morton) - updated based upon comments from Tiziano Ionta. Requested further review/ comments and input. Proposed next step is last call. - No consensus to do RFC 5136 revisions for definition. Errata to be sent by Joseph. Steve Baillargeon spoke with authors and agreed to not reopen the document, and they recommend creating a new standards track document. This would add methods for measuring capacity. -"x"WAMP issues raised at IETF 80 - No follow up from Yakov Stein to this point. There will be TWAMP EAMTC interop event in September in Paris (Those interested in participating, please contact Yakov). Yakov stated that he will create a draft for the next meeting. Yakov will send to the list a link to the EAMTC site. Ongoing work ============ 2. 20' Metrics on the standard track (Al Morton et al) * draft-ietf-ippm-metricstest-03 - YS - Issue with name of this I-D tarball on tools-ietf site - Updated based upon testing begun Feb 2010 - Summary: Ready for WG last call * draft-morton-ippm-testplan-rfc2679-01 - Presentation of actual test results - No correction factors applied to results on slide 12 - Compelling case that both implementations have interpreted the RFC in a statistically equivalent way. Recommended Revisions to RFC 2679 - Revise text in RFC to allow post processing of loss (both implementations can do this, because both store all singletons as required by the RFC) - Drop Type-P-One-Way-Delay-Inverse-Percentile in RFC since neither implementation supported this. - Clarify that lost packets ARE NOT to be included in a conditional distribution of delay measurements (conditioned on successful arrival, which is how most implementations treat the topic) - Adopt as WG document - Perform - Write-up the additional testing for loss (RFC 2680) and delay variation (RFC 3393) (already conducted as part of the completed effort) - Call for additional implementations for interoperability testing (XFO, RIPE, ...) - Matt to check on other implementations 3. 15' Discussion on how to report capacity metrics results. (Al Morton) - draft-ietf-ippm-reporting-metrics-05 - Steve Baillargeon (SB) - Good theoretical approach, but operators sometimes select specific packet sizes and train sizes to determine a "meaningful" performance value. - Agreed that packet size, train size, need to be reported - Yakov Stein (YS) - Need to report the kind of packet (e.g., TCP, UDP). - SB - This draft is about reporting, what is needed is agreement on and documentation of "Method." - YS- Not sure method for measuring capacity. Requested enablement of secret sauce and testing of multiple layers. - DM - Should not define secret sauce things that are not interoperable. Desirable to be able to test Y.1541 IP Network Performance Objective (NPO). - YS - Taking into account existing background flows is challenging. - Next - Finish WG on short-term reporting - Include Barry Constantine comments (if not already addressed, Al to check this) Possible new work ================= 4. 30' TWAMP additons. * Alternative proposal (Al Morton) draft-morton-ippm-twamp-rate-00.txt - SB - Did not see announcement of this draft. Requested that wg mailing list archives be checked. - SB - Mobile operators are asking for more than "Raw" capacity, asking for available capacity across multiple paths. - SB - Description of methods on slide 4 is incomplete. Methods are grouping, variable fixed/variable delta or tau. - AM - No IPR on anything said so far. - YS - Implementing something similar to Burst Measurement. - Session sender decides the rate (Slide 5) - SB - Why choose only line rate. In some environments this may create impairments. Asserted that rate of variation is needed to determine available capacity. - DM - Proposed considering the characteristics of different networks. - SB - Solution should be able to support all networks. * TWAMP Value-Added Octets (Steve Baillargeon/Andreas Johnsson) draft-baillargeon-ippm-twamp-value-added-octets-01 - Presented and described a slide on other capacity measurement methods (Not in slide deck on ietf site) - AM - Disagreed with assertion on slide 2 that TWAMP reflector was not changed. - SB - Disagreed, only extended headers. - AM - Slide 3 - Burst not defining test packets as control packets. Objects to this overall comparison approach and claims there a number of mistakes. Claims that Value Added Octets mode is very complex. - Intrusive is impact that test traffic makes on real traffic. Related to buffer congestion that creates packet loss. YS - Suggested stating minimization of impact to existing streams. New Proposal - Slides 4 - 10 - A response to Al's draft to add forward/reverse and asynchronous mesaurements - These are NOT covered in the subject draft - Slides were placed on ietf.org, meeting materials around 9:30 AM today - AM, DM - Wireline and wireless needs one way capacity measurements - YS - Should ippm be focusing on TWAMP and not OWAMP due to its complexity (e.g., accurate time stamp, sending session schedule, fetching results) - YS - Can two-way mode run without control? SB - Yes. - YS - Forward/ reverse modes are significant change to TWAMP. - MZ - What may actually be needed is a requirements document. - AM - Heroic effort to combine these, but ... wet blanket is IPR declaration. Valuable to keep these separate. - AM - Reflector is complex, it is proposed as optional. Thinks his method is simpler. - SB - Note that core TWAMP specs have IPR declarations. - YS - Vendors like TWAMP reflector simplicity, is implemented on SFPs. Reflector without control protocol is desirable. - Sumita (Svante Ekelin?) Ericsson - Has this been tested in Mobile backhaul? AM - Yes, but not ready to share results. * Discussion on how to continue with this effort. - DM - Agreed with Matt that defining requirements, use cases appears to be best next step. 5. 15' Change status of RFC5136 to standards track. (Steve Baillargeon) - Multiple methods to measure capacity - New document describing general method to test capacity (trains, inter-packet interval, inter-train interval). - SB - Someone from Ericsson could define a new draft. - MZ - Prior description did not appear to be standards track. - MZ - Need a draft outlining what is to be done. - DM - Asked if a draft that specified parameters for an existing method would be standard. MZ - Yes. - DM - Also pointed out that new protocols and methods are being proposed here. - AM/MZ - Can (some) methods be standardized. Narrow down the set of choices. Most of the more successful methods are based on Stream-based Rate Methods. - YS - Note that some methods require dynamic changes of parameters based upon feedback (e.g., detecting increase of delay). - AM - Searching for Available Capacity is more complex than validating a certain value. of capacity is present and "available for the user to utilize". 6. 30' Future of the group - Chairs' slides 11,12 - DM - mentioned proposed Cross Stratum Optimization (CSO) IRTF group and Software-Defined Network Protocol (SDNP) Bar Bof - Will send out information to ippm mailing list. - AM - Measurements of mobile networks working group may have additional parameters, considerations. - YS - OWAMP ambiguous, criticism. Henk requested Yakov to share link. - AM - Pull out referenced sections/ideas (not necessarily wording) from OWAMP and create a complete TWAMP spec independent of OWAMP for readability sake. - MZ - Need to get more people working on this. 7. AOB