MOPs November 21 2019 - First MOPs WG meeting ever! Alas, in conflict with TSVWG... Chairs: Leslie Daigle, Glenn Deen No comments on agenda bash Intro to WG Review of Concrete work items [15min] Draft of edge network operational considerations for streaming media - Jake Holland - draft-jholland-mops-taxonomy On milestone but not adopted Key question: how to calculate the demand vs capacity in a way that will be useful for communicating today and archiving for posterity [Jake] there is a mismatch between the maximum [Martin Thompson] try to break to where the capacity is important (last mile, core) [Dave Oran] is there a correlation between high demand and live data. [Glenn Deen] perhaps produce two charts, one for on-demand file to stream delivery and the other for live. different metrics are important, latency, BW,.. [Jake] there is some trext about ABR in the document [Leslie Daigle] it seems there is interset in the content, need to see if this is a WG asopt. Action [Leslie] are we in favor of adopting [Leslie] Some hum for adopting low hum against it. 3 for in the jabber. [Aaron Falk] use known metrics. [Jake] will be happy to have co-authors. Hum taken to adopt the document - hum appeared to support adoption ### Updates from elsewhere * Updates from other organizations' work - [15min] Sanjay Mishra - Streaming Video Alliance Labs initiative slide 3 is the list of the technical WGs SVA labs open source slide 6 what they did so far in open caching [Leslie] did you provide feedback to the CDNI WG, not only brining new work. [Sanjay] the WG chiar are aware of what is happening in SVA. [Aaron] the slide colors are bad no more comments and questions - [5min] Glenn Deen -- reflections from SMPTE 2019 Time synchronization is important [Roni Even]How to synchronize media from different sources e.g. different cameras with gos clocks. How is the media sent over the IP network and address the clock resolutions [Aaron] there is work in ITU on time solution. it is good if we can capture in MOPS the use case [Dave] data centers have more precise time and should be taken into account. use atomic clocks that are now cheap [Glenn] there is no one way for measuring or distributing time. [Aaron] what is the focus in Jake's draft, production ,...? [Leslie] on the edge side [Glenn] security and integrity of time distribution is important in this space and maybe an new area of work/focus [Aaron] capturing SMPTE operational use cases and issues around time in draft maybe useful ### General discussion of issues * Operational Issues Observed - [15min] Igor Lubashev -- QUIC and streaming how to measure delay and loss when the inforamtion is encryptes (QUIC) [Spencer Dawkins] first time OPS WG discuss this topic and not in Transport. Why using similar TCP flow is not enough [Igor] TCP and UDP applications getting different treatment. [Bernard Aboba] this assumes similar streaming applications, need more detail to allow the network to support for example layered media Spin bit in QUIC V1; support for loss bits may be in a QUIC extension. Need feedback from operators about the need for such montoring tools, were not present in QUIC WG [Spencer] explict signalling does it involve the endpoint [Igor] yes. [Spencer] do not understand how you can measure without involving the endpoints. This is a good direction, wish we started years ago [Igor] QUIC is new and scaled up the problem [Colin Perkins] there is a draft in tsvwg about the effects of header encryptions, please review the part about the network managment [Sanjay] this is good, there was feedback from operators and will continue to follow [Saker] how do you do without involving endpoints, there is work in MOSQUE to add explicit signals using proxies [Emile Stephan] as operator require the monitoring of network, were involved in spin bit and now working on loss bit. [Spencer] there was the plus BOF in Berlin talkin about providing information to on path observers which could not form a WG. [Emile] do not provide the session keys to the observers. endpoint support is important [Colin] in conferencing video there is the RTCP for reporting ftom the Endpoint. using reports from endpoints can be helpful. General discussion of issues [Eric Vyncke, responsible AD] happy with the session AOB Thanks to Roni Even for taking notes during the session. High level summary as posted to Operations & Management wiki: First meeting as a WG. Discussion of draft-jholland-mops-taxonomy (edge network operational considerations for streaming media), and formal adoption as a WG work item. Jake Holland would welcome a co-author to help move it forward. Updates from elsewhere -- Streaming Video Alliance's use of CDNI protocol work; Society of Motion Picture and Television Engineers (SMPTE) recent meeting highlights -- clear dependencies on reliable and granular time reporting in operations. Issues for streaming operations -- overview of QUIC implications for networks streaming video. There are extensions proposed in the QUIC WG, and operators interested in media delivery should go there to weigh in.