IETF
iccrg
iccrg@jabber.ietf.org
Monday, November 8, 2021< ^ >
jishac has set the subject to: Internet Congestion Control Research Group
Room Configuration
Room Occupants

GMT+0
[06:20:46] Glen joins the room
[06:20:50] Glen leaves the room
[15:45:03] Praveen Balasubramanian_web_831 joins the room
[15:45:03] Lin Han_web_785 joins the room
[15:45:03] Paolo Saviano_web_832 joins the room
[15:45:03] Daniel Havey_web_148 joins the room
[15:45:03] Frédéric Fieau_web_206 joins the room
[15:45:03] Nicolas Kuhn_web_930 joins the room
[15:45:03] Mazen Khaddam_web_555 joins the room
[15:45:37] Lin Han_web_785 leaves the room
[15:46:36] Jonathan Morton_web_781 joins the room
[15:47:54] Nick Banks_web_255 joins the room
[15:48:21] Gorry Fairhurst_web_742 joins the room
[15:48:49] Gorry Fairhurst_web_742 leaves the room
[15:48:53] Gorry Fairhurst_web_577 joins the room
[15:49:38] Frode Kileng_web_430 joins the room
[15:49:58] Steffen Klassert_web_630 joins the room
[15:50:53] frodek joins the room
[15:50:56] Mathis Engelbart_web_246 joins the room
[15:52:52] Tom Jones_web_719 joins the room
[15:52:53] Jana Iyengar_web_296 joins the room
[15:53:21] Jeongkeun Lee_web_358 joins the room
[15:53:46] Stephan Emile_web_877 joins the room
[15:53:49] Mazen Khaddam_web_555 leaves the room
[15:53:50] Stephan Emile_web_877 leaves the room
[15:53:51] Stephan Emile_web_692 joins the room
[15:54:13] Richard Scheffenegger_web_968 joins the room
[15:54:20] Mazen Khaddam_web_516 joins the room
[15:54:36] Martin Horneffer_web_150 joins the room
[15:54:50] Magnus Westerlund_web_268 joins the room
[15:54:52] Michael Tüxen_web_682 joins the room
[15:55:54] Zaheduzzaman Sarker_web_944 joins the room
[15:56:16] Hirochika Asai_web_306 joins the room
[15:56:16] David Millman_web_983 joins the room
[15:56:17] Kazuaki Ueda_web_949 joins the room
[15:56:21] Meetecho joins the room
[15:56:57] Neal Cardwell_web_814 joins the room
[15:57:14] John Border_web_618 joins the room
[15:57:22] Ingemar Johansson_web_807 joins the room
[15:57:27] Alessandro Ghedini_web_387 joins the room
[15:57:32] Juhamatti Kuusisaari_web_332 joins the room
[15:58:16] Jake Holland_web_276 joins the room
[15:58:23] Markus Amend_web_734 joins the room
[15:58:24] Mirja Kühlewind_web_924 joins the room
[15:58:43] Markus Amend_web_734 leaves the room
[15:58:45] Renan Krishna_web_744 joins the room
[15:58:47] Markus Amend_web_245 joins the room
[15:58:56] Martin Duke_web_661 joins the room
[15:59:05] Eric Rosenberg_web_116 joins the room
[15:59:20] <Jake Holland_web_276> I can help with notes, sure.
[15:59:24] Nathalie Moreno_web_659 joins the room
[15:59:27] Anna Brunstrom_web_276 joins the room
[15:59:29] Tianji Jiang_web_637 joins the room
[15:59:33] Bhavit Shah_web_639 joins the room
[15:59:38] Xuesong Geng_web_122 joins the room
[15:59:49] Theresa Enghardt_web_720 joins the room
[15:59:55] Lucas Pardue_web_267 joins the room
[16:00:07] <Lucas Pardue_web_267> i can scribe
[16:00:10] Eric Kinnear_web_876 joins the room
[16:00:16] Marcus Ihlar_web_605 joins the room
[16:00:20] Stuart Cheshire_web_101 joins the room
[16:00:33] Ayush Mishra_web_874 joins the room
[16:00:39] Joseph Beshay_web_989 joins the room
[16:00:46] Paul Congdon_web_951 joins the room
[16:00:51] <Lucas Pardue_web_267> please prepend your comment with "MIC:" if you would like be to relay a question to the microphone
[16:01:06] Rodney Grimes_web_929 joins the room
[16:01:13] Hajime Fujita_web_925 joins the room
[16:01:19] Spencer Dawkins_web_947 joins the room
[16:01:23] Mo Zanaty_web_319 joins the room
[16:01:32] Roland Bless_web_438 joins the room
[16:01:42] Anant Shah_web_802 joins the room
[16:01:46] Kyle Rose_web_178 joins the room
[16:01:46] Lin Han_web_136 joins the room
[16:02:03] Ian Swett_web_910 joins the room
[16:02:22] Joerg Ott_web_989 joins the room
[16:02:23] John Kaippallimalil_web_601 joins the room
[16:02:25] Bhavit Shah_web_639 leaves the room
[16:02:27] Rohit Abhishek_web_508 joins the room
[16:02:29] Bhavit Shah_web_269 joins the room
[16:02:34] Alexander Azimov_web_714 joins the room
[16:02:34] Mihail Zverev_web_764 joins the room
[16:03:09] Luis Contreras_web_391 joins the room
[16:03:44] Michael Welzl_web_775 joins the room
[16:03:52] Philip Eardley_web_285 joins the room
[16:04:50] Xavier de Foy_web_227 joins the room
[16:04:57] Ana Custura_web_286 joins the room
[16:07:09] Matt Joras_web_213 joins the room
[16:07:22] David Baldassin_web_109 joins the room
[16:07:55] Matt Joras_web_213 leaves the room
[16:07:59] Matt Joras_web_106 joins the room
[16:08:45] Youngkwon Lim_web_828 joins the room
[16:09:15] Matt Joras_web_106 leaves the room
[16:09:19] Matt Joras_web_850 joins the room
[16:09:29] <Jonathan Morton_web_781> it is still an O(log N) mechanism
[16:10:50] Matt Joras_web_850 leaves the room
[16:10:54] Matt Joras_web_351 joins the room
[16:13:00] Lixia Zhang_web_847 joins the room
[16:15:00] Youngkwon Lim_web_828 leaves the room
[16:15:21] Youngkwon Lim_web_130 joins the room
[16:15:39] Przemysław Król_web_663 joins the room
[16:15:47] Paul Bottorff_web_522 joins the room
[16:19:25] Przemysław Król_web_663 leaves the room
[16:20:20] Koen De Schepper_web_205 joins the room
[16:21:34] Koen De Schepper_web_205 leaves the room
[16:21:38] Koen De Schepper_web_773 joins the room
[16:22:08] Koen De Schepper_web_773 leaves the room
[16:22:12] Koen De Schepper_web_757 joins the room
[16:23:00] Shicheng Wang_web_945 joins the room
[16:24:00] Dragana Damjanovic_web_129 joins the room
[16:30:33] Mat Ford_web_679 joins the room
[16:32:48] Dragana Damjanovic_web_129 leaves the room
[16:35:19] <Lin Han_web_136> can this algorithm be used for congestions happing in multi hops in Internet?
[16:35:35] Ike Kunze_web_506 joins the room
[16:36:17] <Jonathan Morton_web_781> that seems difficult, it seems to rely on a single admin domain
[16:36:50] <Martin Duke_web_661> how many agenda items are we going to get through today?
[16:37:25] <Lin Han_web_136> if ignoring the admin issue? @Jonathan
[16:37:56] <Jonathan Morton_web_781> some of the signalling packets operate at L3, so perhaps in theory
[16:37:56] Tim Chown_web_207 joins the room
[16:38:03] <Jonathan Morton_web_781> maybe ask that on list
[16:38:07] Xuesong Geng_web_122 leaves the room
[16:38:15] <Lin Han_web_136> Thx
[16:38:19] Shicheng Wang_web_945 leaves the room
[16:40:23] <Gorry Fairhurst_web_577> It will be interesting to see how other BBR-basd
[16:40:39] <Jonathan Morton_web_781> or discuss in Gather, if you and Mr. Lee are there
[16:40:41] Ana Custura_web_286 leaves the room
[16:40:52] Rohit Abhishek_web_508 leaves the room
[16:41:07] Mat Ford_web_679 leaves the room
[16:41:39] David Baldassin_web_109 leaves the room
[16:41:43] Dragana Damjanovic_web_474 joins the room
[16:43:13] Koen De Schepper_web_757 leaves the room
[16:43:16] <Jeongkeun Lee_web_358> Thanks a lot for the comments/questions to SFC. Re: Lin Han's question, applying SFC to Internet should be possible if we can mitigate the security concern. As I understand, BBR is making use of ECN markings done by Internet or DOCSIS routers based on a certain trust model. If ICCRG is interested in, we can further discuss.
[16:43:27] <Gorry Fairhurst_web_577> ...based work to extend DCCP now develops, and whether BBR settles on a specific specification that might be captured as a DCCP CCID standards document.
[16:43:48] Yoshifumi Nishida_web_407 joins the room
[16:46:53] Markus Amend_web_245 leaves the room
[16:46:57] Markus Amend_web_731 joins the room
[16:47:09] Markus Amend_web_731 leaves the room
[16:47:10] David Millman_web_983 leaves the room
[16:47:13] Markus Amend_web_664 joins the room
[16:47:14] David Millman_web_736 joins the room
[16:47:47] <Jake Holland_web_276> Gorry, your sound was low.  Was that "please discuss on either" or "please discuss on tsvwg"?
[16:48:00] David Black_web_155 joins the room
[16:48:25] <Gorry Fairhurst_web_577> Discuss either place, but discuss "algorithms and BBR issues on ICCRG"
[16:50:21] David Black_web_155 leaves the room
[16:50:25] David Black_web_170 joins the room
[16:50:43] David Black_web_170 leaves the room
[16:50:47] David Black_web_454 joins the room
[16:50:49] Dragana Damjanovic_web_474 leaves the room
[16:52:11] Matt Joras_web_351 leaves the room
[16:52:15] Matt Joras_web_732 joins the room
[16:52:45] Lijun Dong_web_602 joins the room
[16:52:47] Tianji Jiang_web_637 leaves the room
[16:52:51] Tianji Jiang_web_478 joins the room
[16:53:34] <Lucas Pardue_web_267> dumb question: how does one determine the server's CC algorithm?
[16:53:58] Craig Taylor_web_833 joins the room
[16:54:07] <Jonathan Morton_web_781> I think reading the "census" paper would explain that
[16:55:16] <Jake Holland_web_276> https://www.comp.nus.edu.sg/~~ayush/images/sigmetrics2020-gordon.pdf
[16:55:36] <Lucas Pardue_web_267> thanks Jake
[16:56:56] <Jana Iyengar_web_296> @lucas, basically by emulating specific network loss/delay behavior at the client and observing sender's response. (at its core, this is how the SIGMETRICS paper's techniques work)
[16:58:10] <Lucas Pardue_web_267> ok, sounds like something that clients would not run during real connections
[16:58:10] <Jake Holland_web_276> (i'll comment that the akamai observations in that paper are likely to be out of date today btw, but were very interesting to read :) .)
[16:58:12] Alexander Azimov_web_714 leaves the room
[16:58:47] Alexander Azimov_web_800 joins the room
[16:59:06] <Roland Bless_web_438> There are probably some devices out there that cannot be "upgraded" to BBR, so they have no choice to switch their CC
[16:59:34] <Jana Iyengar_web_296> @lucas - Yup, these are active measurements, meant to probe the congestion controller.
[17:00:19] <Gorry Fairhurst_web_577> So the metric suggested to maximise is "throughput", I wonder how this works-out when you also have flows that wish to minimise collateral loss/latency, rather than increase their share of capacity.
[17:01:18] Dragana Damjanovic_web_285 joins the room
[17:01:36] <Anant Shah_web_802> I should probably dive into the paper more, but I'm curious what the variety of type of flows was in this study? Large video delivery along with small but many web objects etc. A very typical scenario CDNs have to consider
[17:02:05] <Martin Duke_web_661> but if the slope is negative, this does not hold, yes?
[17:02:27] <Martin Duke_web_661> nope, never mind
[17:02:50] Greg White_web_579 joins the room
[17:03:05] <Jana Iyengar_web_296> @Anant: I believe this analysis does not use real traffic profiles, but uses long running flows
[17:03:21] <Anant Shah_web_802> Ack. Thanks Jana
[17:04:28] Geoff Huston_web_949 joins the room
[17:05:21] Dragana Damjanovic_web_285 leaves the room
[17:05:42] <Martin Duke_web_661> Is RTT = min RTT?
[17:06:25] Geoff Huston_web_949 leaves the room
[17:06:49] <Jonathan Morton_web_781> I think it is native RTT, without queueing
[17:07:17] <Jake Holland_web_276> it's pretty hard to compare congestion controller behavior across short flows.  How to approach that issue would be a good research topic; there's only so much a connection-based controller can do...
[17:07:30] <Jonathan Morton_web_781> in a shared queue, the induced delay is experienced by all flows in addition to the native
[17:07:32] Gabriel Montenegro_web_945 joins the room
[17:07:37] Steffen Klassert_web_630 leaves the room
[17:08:00] <Jonathan Morton_web_781> (and it's dynamic)
[17:08:47] <Martin Duke_web_661> So IIUC, people adopt BBR until the ambient buffers small, at which point Cubic is fine for whomever is left on Cubic
[17:09:50] <Jonathan Morton_web_781> I understood it as: short paths are more likely to prefer CUBIC over BBR, long paths to prefer BBR over CUBIC, in a heterogeneous environment
[17:10:40] <Roland Bless_web_438> Yes, because Cubic has advantages for short RTTs, whereas BBR has advantages for large RTTs...
[17:10:59] <Roland Bless_web_438> w.r.t. to their share
[17:11:00] <Jake Holland_web_276> i thought the chart about depending on queue size said deep queues favor cubic when optimizing for throughput, which i think makes sense.  it'll grow bigger before falling off.
[17:11:08] Jafar Shafaeddin_web_270 joins the room
[17:11:59] Yuchung Cheng_web_619 joins the room
[17:13:55] Jafar Shafaeddin_web_270 leaves the room
[17:14:17] Mazen Khaddam_web_516 leaves the room
[17:14:21] Mazen Khaddam_web_538 joins the room
[17:14:44] <Roland Bless_web_438> @Jake that's also true in case the flows have the same RTT
[17:15:09] <Anant Shah_web_802> One more thing, we gotta start thinking about QoE (rebuffering etc) when evaluating ccalgos not just throughput. But that's a broader comment not just for this talk.
[17:15:18] <Lucas Pardue_web_267> is there any research that has measured connection duration on the Internet / Web?
[17:16:09] Mazen Khaddam_web_538 leaves the room
[17:16:54] <Theresa Enghardt_web_720> The most recent one should be this one: https://dl.acm.org/doi/10.1145/3472305.3472321 (ANRW 2021)
[17:19:04] Geraldo Júnior_web_974 joins the room
[17:20:38] <Lucas Pardue_web_267> good call Theresa, thanks! Video at https://irtf.org/anrw/2021/4-ANRW2021-59.m4v
[17:21:39] Vidhi Goel_web_479 joins the room
[17:21:42] <Nicolas Kuhn_web_930> On the flow sizes, forward and return ratios, i found this pointer very useful : M. Trevisan, D. Giordano, I. Drago, M. M. Munafò and M. Mellia, "Five Years at the Edge: Watching Internet From the ISP Network," in IEEE/ACM Transactions on Networking, vol. 28, no. 2, pp. 561-574, April 2020, doi: 10.1109/TNET.2020.2967588.
[17:22:41] <Martin Duke_web_661> IMO the real tipping point is when the research community recommends that the main stacks switch their default
[17:23:09] <Jonathan Morton_web_781> that is probably important in practice, yes
[17:23:19] <Jake Holland_web_276> the real tipping point is when those stacks ship it :)
[17:23:23] <Jonathan Morton_web_781> but those recommendations will take into account research of this type
[17:23:40] <Jake Holland_web_276> unless of course "main stacks" is a misnomer in the brave new world of quic stacks shipped with apps...
[17:23:52] <Martin Duke_web_661> true
[17:24:40] Lixia Zhang_web_847 leaves the room
[17:24:44] Lixia Zhang_web_928 joins the room
[17:24:45] <Jonathan Morton_web_781> "shrink" -> "stop extending right edge"
[17:25:15] <Jake Holland_web_276> "we will live with a zoo of algorithms indefinitely" seems like a reasonable conclusion.
[17:25:44] <Martin Duke_web_661> we still have a ton of customers on HSTCP :grimacing:
[17:25:48] <Jonathan Morton_web_781> which in turn implies that new CC work must coexist reasonably with all of them
[17:26:41] <Jake Holland_web_276> either that or "some kind of policing should probably be rolled out", c.f. huitema at 105 iirc (in was it tsvarea?)
[17:27:47] <Michael Welzl_web_775> the return of conex?
[17:28:02] <Michael Welzl_web_775> by then, there's also ipv6 everywhere, so that fits
[17:28:21] Rodney Grimes_web_929 leaves the room
[17:28:49] <Spencer Dawkins_web_947> @jake - I think you mean https://datatracker.ietf.org/meeting/105/materials/slides-105-tsvarea-congestion-defense-in-depth-00
[17:29:16] <Jake Holland_web_276> yes
[17:30:52] <Michael Welzl_web_775> but there are also incentives to deploy e.g. the thing we're listening to right now. not all flows are capacity-seeking (actually, not many really are, I'd say)
[17:32:21] <Martin Duke_web_661> Welcome to my world, Praveen
[17:32:27] <Spencer Dawkins_web_947> @Michael, I think once you say flows are application-limited (that's what we're talking about, perhaps?), all the transport researchers explode in a puff of smoke.
[17:32:45] <Jana Iyengar_web_296> we should turn those questions and answers between you all into a draft :-)
[17:33:05] <Michael Welzl_web_775> @Spencer: not *all*  :-D  I still exist     :-D
[17:33:53] <Spencer Dawkins_web_947> @Jana, I agree, and I think that's true of many of the ICCRG meetings I've attended on Meetecho!
[17:33:58] <Jonathan Morton_web_781> it might be accurate to say that many flows are too short for their congestion control to reach available capacity
[17:34:26] <Michael Welzl_web_775> but seriously yes, the research focus (for the Internet, not DC, there it may be all right) should shift at some point, to be more in line with the reality of flow lengths and application-limited transfers
[17:34:36] <Spencer Dawkins_web_947> Who's holding a pen for that? Or just setting up a github repo? I'd contribute ...
[17:34:43] <Neal Cardwell_web_814> The "• CUBIC dominates BBRv2 across a range of test cases" is a known issue we are working on
[17:35:27] <Jake Holland_web_276> yeah, that's always been a hard problem.
[17:35:27] <Jonathan Morton_web_781> the slow-start phase is probably an under-researched topic
[17:35:34] <Matt Joras_web_732> I've long thought that the reason researchers focus on not-application-limited workflows is for the really simple reason that those are easiest to emulate.It's very hard to emulate "realistic" application workflows, especially when you're an independent researcher that can't easily sample real workflows.
[17:36:03] <Neal Cardwell_web_814> Just wanted to mention that we have a "fast path" implementation for BBRv2 that takes care of the CPU usage issues, at least in our environment.
[17:36:56] <Spencer Dawkins_web_947> @Jonathan - I'd agree - how far you get into slow start would also matter, and all of the things that go along with that.
[17:37:11] <Ayush Mishra_web_874> Hi everyone, thank you for all the comments! I will address some of the main concerns in the chat here:
@Gorry: In the paper, we assume that the senders are only trying to maximize their throughput. We did this to keep things simple in the paper, but this is obviously not ideal. In our complete model, we will be considering both throughput and queueing delay.
@Martin: Yes, the NE observation does not hold if the gradient at the intersection point is negative. In general, we have observed that the intersection point always has a positive curve, since adding more BBR flows can never _reduce_ the total bandwidth received by the BBR flows.
@Anant: The workloads in these experiments are toy workloads (long flows). We need to look into different flow sizes, workloads and utility functions.
@Theresa, @Nicholas thanks for the resources!
Please feel free to reach out to me at ayush@comp.nus.edu.sg!
PS: @Jake: Noted! I guess it's time to fire up our measurement tool again and see what you guys are up to ;)
PPS: The zoo is only going to get worse. We have also been looking into QUIC implementations of BBR, CUBIC and BBRv2 - and there is already a fair amount a speciation in these implementations.
[17:38:12] <Ayush Mishra_web_874> #TIL the meetecho chat does not format well :(
[17:38:45] <Jonathan Morton_web_781> putting each comment in a separate message might have worked better
[17:39:00] <Jonathan Morton_web_781> I think we can read this though
[17:39:02] <Spencer Dawkins_web_947> Did @jJana say "on the channel"? Is that the mailing list, or in slack?
[17:39:15] <Ayush Mishra_web_874> @Jonathan Yes, I should've done that
[17:39:35] <Jana Iyengar_web_296> @Spencer -- I meant on email, at iccrg@irtf.org
[17:40:13] <Spencer Dawkins_web_947> @Jana - thanks! I'm still discovering what's on Slack at the IETF ... :grin:
[17:40:18] <Martin Duke_web_661> So is v1 = the -00 draft?
[17:40:19] <Martin Duke_web_661> ish?
[17:40:36] <Praveen Balasubramanian_web_831> Hi Vidhi, I will follow up on your question about GAIN factor offline.
[17:41:04] <Jana Iyengar_web_296> yup, I wouldn't mind doing an ICCRG slack channel if there's interest!
[17:41:19] <Vidhi Goel_web_479> +1 for ICCRG slack channel
[17:41:29] <Roland Bless_web_438> @Martin: yes
[17:41:30] Gabriel Montenegro_web_945 leaves the room
[17:41:35] Michio Honda_web_627 joins the room
[17:41:35] <Spencer Dawkins_web_947> @Ayush Mishra - providing good content makes it worth us decoding the formatting ...
[17:44:02] <Michael Welzl_web_775> I'm happy that the authors updated the BBR draft. This is so important for this community!
[17:44:08] <Lucas Pardue_web_267> this could be an SVG in a doc right?? :wink:
[17:45:32] Mazen Khaddam_web_167 joins the room
[17:45:41] <Lucas Pardue_web_267> that probeRTT trough is a good viz
[17:46:02] <Spencer Dawkins_web_947> @Lucas - yes. I think there are I-Ds with SVG already.
[17:46:07] <Kyle Rose_web_178> Is there an out-of-tree kernel module source package/repo?
[17:46:46] <Lucas Pardue_web_267> https://github.com/martinthomson/aasvg - Convert ASCII art diagrams into SVG.
[17:46:53] <Praveen Balasubramanian_web_831> yes Kyle, that's the one we used as reference. https://github.com/google/bbr/blob/v2alpha/net/ipv4/tcp_bbr2.c
[17:47:01] <Kyle Rose_web_178> Great!
[17:48:15] <Neal Cardwell_web_814> Re "So is v1 = the -00 draft?", yes, the -00 draft is BBRv1. And the -01 draft is the first cut at describing BBRv2. We will expand ASAP.
[17:49:10] <Jana Iyengar_web_296> FYI -- I've created a #iccrg channel inside the IETF slack workspace (ietf.slack.com) Please join!
[17:49:18] Chi-Jiun Su_web_122 joins the room
[17:49:18] Geoff Huston_web_371 joins the room
[17:49:28] Stephan Emile_web_692 leaves the room
[17:50:01] <Neal Cardwell_web_814> Yes, we have a patch for that one on the TCP side and plan to use ASAP.
[17:50:33] Michael Tüxen_web_682 leaves the room
[17:50:37] Michael Tüxen_web_670 joins the room
[17:50:45] Michael Tüxen_web_670 leaves the room
[17:50:49] Michael Tüxen_web_659 joins the room
[17:51:14] Anant Shah_web_802 leaves the room
[17:51:21] Mazen Khaddam_web_167 leaves the room
[17:52:01] Xavier de Foy_web_227 leaves the room
[17:55:24] <Jeongkeun Lee_web_358> Are you considering ECN for both WAN and DC-internal use cases?
[17:56:31] Enke Chen_web_725 joins the room
[17:56:32] Lin Han_web_136 leaves the room
[17:57:16] Frédéric Fieau_web_206 leaves the room
[17:58:22] Enke Chen_web_725 leaves the room
[17:58:55] Hajime Fujita_web_925 leaves the room
[17:59:14] Lijun Dong_web_602 leaves the room
[18:01:10] David Millman_web_736 leaves the room
[18:01:50] Ingemar Johansson_web_807 leaves the room
[18:02:19] Nick Banks_web_255 leaves the room
[18:02:59] Luis Contreras_web_391 leaves the room
[18:03:07] Tim Chown_web_207 leaves the room
[18:03:58] <Lucas Pardue_web_267> this was my first ICCRG session since the last Prague  meeting. Super interesting, thanks all
[18:04:02] Theresa Enghardt_web_720 leaves the room
[18:04:02] <Martin Duke_web_661> Top notch, Jana!
[18:04:03] Philip Eardley_web_285 leaves the room
[18:04:03] <Neal Cardwell_web_814> Thanks, Jana!
[18:04:06] Markus Amend_web_664 leaves the room
[18:04:07] <Tom Jones_web_719> thanks
[18:04:08] Youngkwon Lim_web_130 leaves the room
[18:04:08] Mihail Zverev_web_764 leaves the room
[18:04:09] Nathalie Moreno_web_659 leaves the room
[18:04:12] Martin Horneffer_web_150 leaves the room
[18:04:12] Daniel Havey_web_148 leaves the room
[18:04:14] Chi-Jiun Su_web_122 leaves the room
[18:04:15] Roland Bless_web_438 leaves the room
[18:04:16] Ike Kunze_web_506 leaves the room
[18:04:17] David Black_web_454 leaves the room
[18:04:18] <Anna Brunstrom_web_276> thanks, all, great presentations
[18:04:19] <Michael Welzl_web_775> Great, Jana!  See you all!
[18:04:20] Geraldo Júnior_web_974 leaves the room
[18:04:22] Magnus Westerlund_web_268 leaves the room
[18:04:23] Jeongkeun Lee_web_358 leaves the room
[18:04:23] Geoff Huston_web_371 leaves the room
[18:04:26] Paul Congdon_web_951 leaves the room
[18:04:33] Ian Swett_web_910 leaves the room
[18:04:36] Ayush Mishra_web_874 leaves the room
[18:04:37] <Michael Welzl_web_775> Very interesting presentations indeed!
[18:04:37] John Border_web_618 leaves the room
[18:04:41] Juhamatti Kuusisaari_web_332 leaves the room
[18:04:41] John Kaippallimalil_web_601 leaves the room
[18:04:42] Matt Joras_web_732 leaves the room
[18:04:43] Eric Rosenberg_web_116 leaves the room
[18:04:43] <Neal Cardwell_web_814> Re "Are you considering ECN for both WAN and DC-internal use cases?" Yes, we are currently using it for DC-internal cases but do plan on using it for WAN and public Internet cases.
[18:04:43] Michael Welzl_web_775 leaves the room
[18:04:44] Frode Kileng_web_430 leaves the room
[18:04:44] Bhavit Shah_web_269 leaves the room
[18:04:45] Lixia Zhang_web_928 leaves the room
[18:04:45] Stuart Cheshire_web_101 leaves the room
[18:04:46] Martin Duke_web_661 leaves the room
[18:04:49] Tom Jones_web_719 leaves the room
[18:04:49] Greg White_web_579 leaves the room
[18:04:50] <Jana Iyengar_web_296> Thank you everyone, and see you all in March!
[18:04:50] Nicolas Kuhn_web_930 leaves the room
[18:04:51] Mathis Engelbart_web_246 leaves the room
[18:04:51] Marcus Ihlar_web_605 leaves the room
[18:04:52] Alessandro Ghedini_web_387 leaves the room
[18:04:52] Praveen Balasubramanian_web_831 leaves the room
[18:04:52] Kyle Rose_web_178 leaves the room
[18:04:53] Eric Kinnear_web_876 leaves the room
[18:04:56] Mirja Kühlewind_web_924 leaves the room
[18:04:57] Yuchung Cheng_web_619 leaves the room
[18:04:58] Joerg Ott_web_989 leaves the room
[18:05:03] Renan Krishna_web_744 leaves the room
[18:05:05] Kazuaki Ueda_web_949 leaves the room
[18:05:11] Anna Brunstrom_web_276 leaves the room
[18:05:12] Mo Zanaty_web_319 leaves the room
[18:05:22] Michael Tüxen_web_659 leaves the room
[18:05:28] Joseph Beshay_web_989 leaves the room
[18:05:45] Vidhi Goel_web_479 leaves the room
[18:06:05] Jake Holland_web_276 leaves the room
[18:06:08] Neal Cardwell_web_814 leaves the room
[18:06:08] Spencer Dawkins_web_947 leaves the room
[18:06:12] Meetecho leaves the room
[18:07:29] Jana Iyengar_web_296 leaves the room
[18:07:32] Lucas Pardue_web_267 leaves the room
[18:08:05] Mazen Khaddam_web_150 joins the room
[18:08:06] Paolo Saviano_web_832 leaves the room
[18:08:06] Richard Scheffenegger_web_968 leaves the room
[18:08:06] Jonathan Morton_web_781 leaves the room
[18:08:06] Gorry Fairhurst_web_577 leaves the room
[18:08:06] Hirochika Asai_web_306 leaves the room
[18:08:06] Zaheduzzaman Sarker_web_944 leaves the room
[18:08:06] Tianji Jiang_web_478 leaves the room
[18:08:06] Yoshifumi Nishida_web_407 leaves the room
[18:08:06] Paul Bottorff_web_522 leaves the room
[18:08:06] Craig Taylor_web_833 leaves the room
[18:08:06] Michio Honda_web_627 leaves the room
[18:08:06] Alexander Azimov_web_800 leaves the room
[18:08:09] Mazen Khaddam_web_150 leaves the room
[20:34:40] frodek leaves the room