395 views
 owned this note
# Multicast for the Web IETF 111 Bar BoF ## Attendees Jake Holland, Akamai Kyle Rose, Akamai Spencer Dawkins, Tencent America Sam Hurst, BBC R&D Mike Bishop (Akamai) Mike McBride (Futurewei) Chris Needham (BBC) Ross Finlayson (Live Networks) Gorry Fairhurst (University of Aberdeen, UK) Chris Lenart (Verizon) John Border (Hughes) chi-jiun su (Hughes) ## Intro (Jake) Meeting under note-well 5: Impact of new game releases Want to use L1 broadcast to support multicasting that content * Problem - people getting significant reductions in performance, which Akamai attributes to congestion, much of which results from traffic that is potentially shareable with multicast * Second problem - Jake wants to save 1 percent of global carbon footprint, which he thinks is achievable * Guessing that 40 percent of game and OS traffic into an ISP may be shareable, and that this is a lowball estimate * Caching helps but has limits Jake described some of the tools that can be used if this traffic IS multicast - AMBI, for example Early ISP feedback has been positive * successful lab trials with 5 ISPs, including cable, fiber, xDSL (but no satellite ISPs during the trials) * need more receivers * Web API is critical for live video (smart TVs, mobile devices) * Web nice-to-have for Game/OS/software downloaders * need more content * Extension rejected for Chromium * MUST require encryption for a new web API * Next steps * Trying to use QUIC framing and Alt-svc instead of web-app join (existing proposal in QUIC as starting point) * Get consensus on security requirements (existing proposal from krose) QA: * Richard Bradbury: BBC is positive on this - have done HTTP objects over multicast * worried by API concerns - don't want to change implementations * realize there are barriers to that * great job at creating an ecosystem, now can you hide the moving parts? * Not just video, also downloads, etc. * Desires an approach in which unicast applications can discover (e.g., via Alt-Svc) the existence of a multicast source for content, which could enable generic applications to leverage multicast without requiring any special support to do so. * Jake - prefers an incremental approach rather than trying to get to transparency from the out-set. * Magnus: so you're using object multicast, not classical (packet) multicast? * Jake - These are classic multicast packets. The current prototype reassembles them to objects in the web app layer. I wouldn't rule out that model, but there's an intermediary that makes it harder. * Kyle - if we can get to transparency eventually, that would be awesome, but can't boil the ocean(s) instantaneously, so not starting there * Jake - there are networking considerations with multicast (just because you want multicast, that doesn't mean your network is going to forward them. If a lot of the traffic starts being multicast, operators have decisions to make) * Tianji - 3GPP is working on architecture for multicast/broadcast service (MBS) in 5G * using the control plane, not the data plane * seems like you want the browser to do the work? * Jake - last I saw in the TR, there's a transparent mode that allows IETF-style group joins, and another mode that uses the northbound interface. Don't object to using the northbound interface, but prefer to use transparent mode. * Tianji - the difference is that the northbound interface is using multicast in the RAN * Richard - there's also discussion about using IGMP for legacy devices that gets translated to control plane signaling at residential gateways. thats MBSF and MBSTF in 5G for multicast packets * Tianji provided document numbers for 5G specifications about this: 3gpp documents: TR 23.247 and TR 23.757 for 5GMBS * Shuai provided https://www.etsi.org/deliver/etsi_ts/126300_126399/126347/16.03.01_60/ts_126347v160301p.pdf from ETSI * Jake: is this hopeless? * Pursue and be positive, but where will this work be? * MBONED, WebTransport, new working group, SECDISPATCH or a BOF next time unless WebTransport is interested (and that would require a recharter) * Spencer after the call was breaking up: DREAM BIG, JAKE! You're not alone ... ## Webex Chat Notes from Carsten Bormann to everyone: 2:03 PM Maybe have a https://codimd.ietf.org/notes-ietf-111-multicast-side-meeting?edit ? from Spencer to everyone: 2:05 PM Oops - not seeing a nink for notes? from Kyle Rose to everyone: 2:05 PM https://codimd.ietf.org/notes-ietf-111-multicast-side-meeting?edit from Spencer to everyone: 2:08 PM Thank you, Kyle! from Spencer to everyone: 2:08 PM I should also ask if Jake wants people to record attendance there from MAPRG Research Group to everyone: 2:10 PM But that's only traffic that is consumed simultaneously by multiple users, no? from shuai zhao to everyone: 2:11 PM Is climate impact mainly caused by power usage? from shuai zhao to everyone: 2:12 PM Thx from Carsten Bormann to everyone: 2:12 PM "Internet" probably includes Bitcoin mining... from Carsten Bormann to everyone: 2:13 PM I think the goal is laudable anyway from Richard Bradbury BBC to everyone: 2:14 PM Some colleagues of mine at BBC R&D are looking at the environmental impact of the Internet. My guess would be that a router/switch has a high base load to power its optics regardless of traffic levels. There may be a small extra load that is proportional to load (because the switching fabric has to think a bit harder when processing packets). from Spencer to everyone: 2:15 PM I'm planning to copy the Webex chat into the notes - any objections? from Carsten Bormann to everyone: 2:16 PM Please do from Spencer to everyone: 2:17 PM Please record your attendance there - we have 33 attendees and 6 names on the attendees list from Kyle Rose to everyone: 2:17 PM Link again: https://akamai.webex.com/akamai-en/url.php?frompanel=false&gourl=https%3A%2F%2Fcodimd.ietf.org%2Fnotes-ietf-111-multicast-side-meeting%3Fedit from Kyle Rose to everyone: 2:17 PM Whoops: https://codimd.ietf.org/notes-ietf-111-multicast-side-meeting?both from John Border to everyone: 2:21 PM Were any satellite ISPs asked? from Gorry Fairhurst to everyone: 2:22 PM GEO with Spot beam satellite s also interesting - there's work in DVB on IP-based Video transmisison within DV. from Mike Bishop (internal) to everyone: 2:23 PM If you're not Jake, please mute. from John Border to everyone: 2:24 PM A big issue with encryption happens if you need to revoke someone's ability to receive. from Kyle Rose to everyone: 2:24 PM Jake, if you make me co-host, I can mute folks who are echoing from Jake Holland (internal) to everyone: 2:26 PM i don't know how to make you co-host, but i'll try... from Kyle Rose to everyone: 2:27 PM You click on the ... next to my name in the participant list and choose "make co-host", IIRC from Mike McBride to everyone: 2:28 PM What's the status of your draft Richard? from Spencer to everyone: 2:28 PM Is there a TAPS participant on the call? :D from Carsten Bormann to everyone: 2:31 PM Video consumption is a bit more likely to be approximately synchronized than software download. from Spencer to everyone: 2:32 PM TO ALL - if you're talking, please check anything i type in the notes with your name next to it. I'm not doing an awesome job from Jake Holland (internal) to everyone: 2:43 PM @Carsten, we have a software downloader prototype that can get concurrency with unsynchronized joins and receiver-driven speeds at the same time. from Jake Holland (internal) to everyone: 2:43 PM we're evaluating it with some of our customers from Spencer to everyone: 2:43 PM Richard - thank you for that observation! I wasn't sure whether that was just me :D from shuai zhao to everyone: 2:44 PM https://www.etsi.org/deliver/etsi_ts/126300_126399/126347/16.03.01_60/ts_126347v160301p.pdf from MAPRG Research Group to everyone: 2:45 PM can we wrap-up at least 5 minutes before the hour to get a bit of a break...? from shuai zhao to everyone: 2:45 PM I believe in 3gpp SA4 WG, they should have those API defined for 5G to make sure MBMS will work for applications from Tianji to everyone: 2:46 PM 3gpp documents: TR 23.247 and TR 23.757 for 5GMBS from Spencer to everyone: 2:46 PM @MAPRG +1 on breaking before the hfrom Carsten Bormann to everyone: 2:03 PM Maybe have a https://codimd.ietf.org/notes-ietf-111-multicast-side-meeting?edit ? from Spencer to everyone: 2:05 PM Oops - not seeing a nink for notes? from Kyle Rose to everyone: 2:05 PM https://codimd.ietf.org/notes-ietf-111-multicast-side-meeting?edit from Spencer to everyone: 2:08 PM Thank you, Kyle! from Spencer to everyone: 2:08 PM I should also ask if Jake wants people to record attendance there from MAPRG Research Group to everyone: 2:10 PM But that's only traffic that is consumed simultaneously by multiple users, no? from shuai zhao to everyone: 2:11 PM Is climate impact mainly caused by power usage? from shuai zhao to everyone: 2:12 PM Thx from Carsten Bormann to everyone: 2:12 PM "Internet" probably includes Bitcoin mining... from Carsten Bormann to everyone: 2:13 PM I think the goal is laudable anyway from Richard Bradbury BBC to everyone: 2:14 PM Some colleagues of mine at BBC R&D are looking at the environmental impact of the Internet. My guess would be that a router/switch has a high base load to power its optics regardless of traffic levels. There may be a small extra load that is proportional to load (because the switching fabric has to think a bit harder when processing packets). from Spencer to everyone: 2:15 PM I'm planning to copy the Webex chat into the notes - any objections? from Carsten Bormann to everyone: 2:16 PM Please do from Spencer to everyone: 2:17 PM Please record your attendance there - we have 33 attendees and 6 names on the attendees list from Kyle Rose to everyone: 2:17 PM Link again: https://akamai.webex.com/akamai-en/url.php?frompanel=false&gourl=https%3A%2F%2Fcodimd.ietf.org%2Fnotes-ietf-111-multicast-side-meeting%3Fedit from Kyle Rose to everyone: 2:17 PM Whoops: https://codimd.ietf.org/notes-ietf-111-multicast-side-meeting?both from John Border to everyone: 2:21 PM Were any satellite ISPs asked? from Gorry Fairhurst to everyone: 2:22 PM GEO with Spot beam satellite s also interesting - there's work in DVB on IP-based Video transmisison within DV. from Mike Bishop (internal) to everyone: 2:23 PM If you're not Jake, please mute. from John Border to everyone: 2:24 PM A big issue with encryption happens if you need to revoke someone's ability to receive. from Kyle Rose to everyone: 2:24 PM Jake, if you make me co-host, I can mute folks who are echoing from Jake Holland (internal) to everyone: 2:26 PM i don't know how to make you co-host, but i'll try... from Kyle Rose to everyone: 2:27 PM You click on the ... next to my name in the participant list and choose "make co-host", IIRC from Mike McBride to everyone: 2:28 PM What's the status of your draft Richard? from Spencer to everyone: 2:28 PM Is there a TAPS participant on the call? :D from Carsten Bormann to everyone: 2:31 PM Video consumption is a bit more likely to be approximately synchronized than software download. from Spencer to everyone: 2:32 PM TO ALL - if you're talking, please check anything i type in the notes with your name next to it. I'm not doing an awesome job from Jake Holland (internal) to everyone: 2:43 PM @Carsten, we have a software downloader prototype that can get concurrency with unsynchronized joins and receiver-driven speeds at the same time. from Jake Holland (internal) to everyone: 2:43 PM we're evaluating it with some of our customers from Spencer to everyone: 2:43 PM Richard - thank you for that observation! I wasn't sure whether that was just me :D from shuai zhao to everyone: 2:44 PM https://www.etsi.org/deliver/etsi_ts/126300_126399/126347/16.03.01_60/ts_126347v160301p.pdf from MAPRG Research Group to everyone: 2:45 PM can we wrap-up at least 5 minutes before the hour to get a bit of a break...? from shuai zhao to everyone: 2:45 PM I believe in 3gpp SA4 WG, they should have those API defined for 5G to make sure MBMS will work for applications from Tianji to everyone: 2:46 PM 3gpp documents: TR 23.247 and TR 23.757 for 5GMBS from Spencer to everyone: 2:46 PM @MAPRG +1 on breaking before the hour ... from Tianji to everyone: 2:46 PM for rel-17 from shuai zhao to everyone: 2:47 PM Yea, that looks abt right from Spencer to everyone: 2:48 PM O from Spencer to everyone: 2:49 PM I'm actually stopping typing in about 2 minutes (others can continue) from Jake Holland (internal) to everyone: 2:49 PM Thanks Spencer. from Spencer to everyone: 2:49 PM Can someone cut and paste the webex chat into the notes? from MAPRG Research Group to everyone: 2:49 PM This is Mirja Kühlewind our ... from Tianji to everyone: 2:46 PM for rel-17 from shuai zhao to everyone: 2:47 PM Yea, that looks abt right from Spencer to everyone: 2:48 PM O from Spencer to everyone: 2:49 PM I'm actually stopping typing in about 2 minutes (others can continue) from Jake Holland (internal) to everyone: 2:49 PM Thanks Spencer. from Spencer to everyone: 2:49 PM Can someone cut and paste the webex chat into the notes? from MAPRG Research Group to everyone: 2:49 PM This is Mirja Kühlewind from Spencer to everyone: 2:50 PM Thanks for your help, Kyle! from Jake Holland (internal) to everyone: 2:50 PM Yes, thanks Kyle :) from Tianji to everyone: 2:51 PM 3gpp documents: TR 23.247 and TR 23.757 for 5GMBS from mysteryperson to everyone: 2:51 PM keep the hope from MAPRG Research Group to everyone: 2:53 PM webtrans might still be busy with othee stuff... from Spencer to everyone: 2:53 PM DREAM BIG, JAKE!!! from mysteryperson to everyone: 2:53 PM cheers all from Richard Bradbury BBC to everyone: 2:54 PM Thanks! from MAPRG Research Group to everyone: 2:54 PM thx! from Sam Hurst BBC to everyone: 2:54 PM Thanks very much Jake!