16:59:47 #startmeeting Network team meeting, 18th July 2022 16:59:47 Meeting started Mon Jul 18 16:59:47 2022 UTC. The chair is ahf. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:59:47 Useful Commands: #action #agreed #help #info #idea #link #topic. 16:59:50 hello hello, welcome 16:59:54 yello 17:00:03 our pad is @ https://pad.riseup.net/p/tor-netteam-2022.1-keep 17:00:09 o/ 17:00:21 o/ 17:00:25 hihi 17:00:30 o/ 17:00:47 o/ 17:01:05 anybody who have anything to our boards: https://gitlab.torproject.org/groups/tpo/core/-/boards 17:01:17 i think they are in great condition due to all the roadmap ticket shuffling we did 17:01:53 seems legit 17:02:20 dgoulet: i am thinking we skip the release talk given that it's your day #1 right now 17:02:47 yah except 17:03:03 048 will likely be one with HS PoW at the very least hehe 17:03:08 ya 17:03:11 very good 17:03:37 for stuff coming in from other teams, we got tor#40635 from geko a bit ago, but dgoulet is CC'ed on that one and i think that is the most important part of it 17:03:37 Uhm, which one of [tpo/core/debian/tor, tpo/core/tor] did you mean? 17:03:40 ugh 17:03:44 core/tor#40635 17:04:05 we have one announcement: 17:04:07 [2022 07 18] [nickm] nickm will be AFK August 18-August 22. 17:04:37 ok, so this is the last meeting where i am around until the 15th of august 17:05:21 we do quite a few meetings every week: we have these monday meetings that i usually run, they are very "checklist-style followed by free conversation", we have our arti internal meetings on some tuesdays, we have arti public meetings on some wednesdays and we have our thursday meeting 17:06:10 i'd suggest you continue to do these monday meetings, and i have been trying to get micah to run them, but he hasn't gotten back to me yet if it fits into his schedule. if it doesn't, i'd say dgoulet could run them here as fallback since the meeting "checklists" are still very C tor related since a lot of it is coordination with other teams, s61, etc. 17:07:01 i think you should not do the internal arti meetings while i am gone and just do tuesday like there was no meetings. i think it would be OK to continue to run the public arti ones, but i am also OK with dropping those. i'd need someone who can *run* them (press start record, etc.) 17:07:14 ahf: who is going to be facilitating this one? 17:07:18 the thursday meetings i'd say: do them if you have something you want to talk about 17:07:19 gaba: read above 17:07:30 oh, dgoulet :) 17:07:33 but it will probably be less structured 17:07:33 i skipped that 17:07:34 thanks 17:07:36 plan wfm :) 17:07:40 gaba: nope, micah and if micah can't do it then dgoulet 8) 17:07:48 there is a whole flowchart here 8) 17:07:57 hehe 17:07:58 arti folks, what do you prefer on wednesday meetings? 17:07:58 ok 17:08:09 i am inclined to say coordinate via irc and ditch these meetings for now? 17:08:35 I'm OK with that. 17:08:45 We might need to bring things that end up blocked to the Thursday meeting. 17:09:10 no problems from me here 17:09:11 yeah, my hope was that you can do more peer-to-peer interactions here rather than bigger sync points 17:09:17 just don't go full anarchy 17:09:49 if you want me to run them, then anarchy is on the menu ;) 17:09:59 ok, great, so summary is: you have monday meetings at the usual times. no other meetings UNLESS some of you decide to go for the thursday meeting, but they won't be mandatory 17:10:04 mike will run the s61 meetings still 17:10:09 EXCELLENT folks 17:10:13 micah: ahahah <3 17:10:15 oh you are talking wednesdays... for mondays at 17h I can! 17:10:24 it's only monday you need to do anything, micah 17:10:30 +1 17:10:36 and it's all in the pad @ https://pad.riseup.net/p/tor-netteam-2022.1-keep 17:10:45 ok team, we also won't have any 1:1 while i am gone 17:11:19 if you have any kind of questions on roadmap, team interactions, etc. etc. etc., feel free to go to micah who can help you (or point you to someone who can) 17:11:31 i am on signal if there is something where you think i need to be aware of it 17:11:37 ok, any questions here? :o 17:12:57 i take that as a no 17:13:05 mikeperry: wanna do s61? 17:13:08 good 2 minutes of silence does it lol 17:13:10 :D 17:13:11 yeah 17:13:15 ok 17:14:21 so the main thing in s61 Objective 1 is jnewsome's work on the link sharing for the sim, and hiro and jim working on Fast+Guard cutoff analysis support and sims 17:15:03 jnewsome: you were working on a shadow bug that was blocking the link sharing? is that still going on? 17:15:19 mikeperry: the bug is now fixed :-D 17:15:39 over the weekend I ran a baseline @ the latest shadow and tornettools 17:15:47 and another with link sharing (but without congestion control) 17:16:14 theoretically next I should run one with link sharing + congestion control, though I should check with you exactly what configs I should be using 17:16:20 ice 17:16:22 nice* 17:16:38 ok that's great 17:17:16 so far the july data is looking usable for comparison there, but obviously july is not done yet 17:17:36 i briefly glanced over the baseling link sharing results; seems like exit traffic throughput drops a fair bit, which makes sense. https://gitlab.torproject.org/jnewsome/sponsor-61-sims/-/jobs/154940/artifacts/file/public/tornet.plot.pages.pdf 17:17:36 how did the comparison look pre-congestion control? did link sharing make the perf much worse? 17:18:05 yup 17:19:04 affected exit traffic quite a bit 17:19:23 didn't seem to affect HS traffic much; makes sense since I think the clusters were largely exits 17:19:33 ok. I think we can run a congestion control sim before july is over, but yeah I will need to give you a branch for that 17:19:48 interesting. ok 17:20:34 I have just updated the script mikeperry so you have a switch to specify the instance(s) you want to download measurements from 17:20:41 like -i op-de7a 17:20:53 we can probably use hiro's script to compare that link-sharing sim to an unfiltered extract from july 17:21:10 hiro: awesome, thanks. I will give that a try today then 17:22:16 juga: it looks like we're got HTTP POST support on all bwauth httpd now? and I think I saw you had a 40130 branch already? 17:22:25 yes 17:22:27 mikeperry: i'll wait to hear from you abt a congestion control branch to run. do you have something I should run as the baseline comparison in the meantime? 17:22:48 mikeperry: when we would like to be using uploading instead of downloading? 17:23:51 jnewsome: I think at this point we're effectively going to be comparing congestion control to congestion control, using live data. and since july is not done yet, we may just do manual? Or perhaps july 12-$today 17:24:26 mikeperry: I think for the clustering to make sense we need a flooding period 17:25:05 juga: if you can get a friendly bwauth to test, whenever you are ready to do so. though one of them that is not bastet or farahavar would be good, since we only suspect those two to be bottlenecked rn 17:25:42 jnewsome: oh you mean for the network model input, not so much baseline perf comparison 17:26:02 I believe september was the best flooding period we have. I can review specific dates after 17:26:20 mikeperry: ok, we'll start testing in longclaw then 17:26:45 juga: you have log events for that XOFF stream event, yeah? 17:27:05 mikeperry: right; and if we're doing a non-cc baseline then exactly what tor version and consensus params etc. if you want we can follow up outside the meeting 17:27:08 mikeperry: err, i've not checked with upload yet 17:27:16 mikeperry: will do 17:27:57 juga: ok. I can also try to take a quick look over the branch too if you want. I might have some further ideas for stuff to watch for 17:28:12 mikeperry: ok, feel free to do it, thanks 17:28:32 juga: ok, I will follow up after 17:28:38 jnewsome: yeah, let's also follow up after 17:28:42 mikeperry: great 17:29:41 cool 17:29:42 geko is out for O4.. but dgoulet and I will be looking at stuff related to the dos 17:30:51 * juga needs to go afk, ttyl 17:31:03 I think that actually covers it for the meeting. i have a big follow-up list but will get to it :) 17:33:23 nice 17:33:40 okay folks, enjoy your time without a lot of meetings and i hope you get some nice hacking going 17:33:47 i will see you all around the 15th of august :-) 17:33:49 o/ 17:33:52 \o/ 17:33:54 #endmeeting