The web in a box - a next generation web framework for the Perl programming language

IRC log for #mojo, 2016-08-06

| Channels | #mojo index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
02:08 FROGGS joined #mojo
02:28 noganex_ joined #mojo
03:24 kaare joined #mojo
03:35 * sri also learned this week that a site in the alexa top 50 at least partially runs on mojolicious :o
04:21 Janos joined #mojo
04:33 jberger Wow!
04:34 jberger Can you say which one?
04:36 lluad joined #mojo
05:10 sri there are no public references, so i think i'll keep it to myself for now
06:07 itaipu joined #mojo
07:15 Vandal joined #mojo
09:08 Zx3 joined #mojo
09:58 stryx` joined #mojo
10:50 cuechan joined #mojo
11:09 itaipu joined #mojo
12:04 tchaves joined #mojo
12:08 kid51 joined #mojo
12:18 mpapec https://gist.github.com/mpapec/2d89b4421a5820057d4e6093447df05c
12:19 mpapec 72 cores (HT enabled xeon E5-2699 v3 @ 2.30GHz)
12:20 mpapec in reference to https://github.com/kraih/mojo/issues/925
12:20 mpapec perl 5.20.3, latest Mojolicious
12:32 itaipu joined #mojo
12:40 mpapec http://niceperl.blogspot.hr/2016/08/ccxxiv-metacpan-weekly-report.html
13:01 asarch joined #mojo
13:11 kid51 joined #mojo
14:00 Janos joined #mojo
14:22 AirDisa_ joined #mojo
14:30 sri mpapec: afraid that doesn't really help
14:32 sri mpapec: what's needed is results from the multicore-benchmark.sh script
14:32 sri which means slow progression, multiple runs with more and more cores
14:33 sri so we can mage a graph to judge usage of cores
14:33 sri s/mage/make/
14:38 sri mpapec: for 72 cores something like this https://gist.github.com/anonymous/c4e1298da9bc2eef8ffb2fc4674f74d6
15:04 sri i guess ideally you would divide the number of requests by the number of workers
15:38 sri i've posted a better benchmark script https://github.com/kraih/mojo/issues/925#issuecomment-238029327
16:18 cuechan joined #mojo
16:30 mpapec sri: I've posted a comment
16:30 mpapec I believe only munin kicks in now and then
16:31 mpapec hope it helps
16:34 lluad joined #mojo
17:01 sri mpapec: interesting
17:01 sri looks like results are meh when the number of connections per worker gets too low
17:01 sri guess 512 connections or so would be better there
17:08 cuechan joined #mojo
17:09 sri and the curve https://github.com/kraih/mojo/issues/925#issuecomment-238034105
17:12 sri jberger: that +1 might really hurt performance it seems when there's too little connections
17:12 sri wonder how many connections the techempower benchmarks use
17:16 mpapec ok, great
17:16 mpapec tell me if you have some other benchmarks
17:17 mpapec I was thinking about some utility which would tell me what are optimal settings for current machine
17:18 mpapec like number of workers, number of file handles
17:18 mpapec and checks for if they are actually available
17:19 sri on such a big box definitely more connections
17:19 sri CONNECTIONS=512
17:19 mpapec ok, so I change this setting only?
17:20 sri yes
17:20 sri you might even be testing the wrk limits :)
17:20 sri it could need a few more threads too
17:20 mpapec it is written in c :)
17:21 sri but only uses 2 threads by default i believe
17:21 mpapec does it make sense to ramp the wrk number of threads?
17:22 mpapec bbl
17:22 sri not sure, could try a -t 4 or so
17:25 sri anyway, it's nice to see confirmation that we can scale pretty well up to 32 cores at least
17:27 sri hardest problem is load balancing, think a slightly different architecture would make sense there, if anyone ever needs a super high performance version of hypnotoad
17:30 sri as in, it's tough to evenly distribute incoming connections among workers
17:57 PryMar56 joined #mojo
18:14 ajr_ joined #mojo
18:24 Zx3 joined #mojo
18:25 kid51 joined #mojo
18:41 Adura joined #mojo
19:11 itaipu joined #mojo
19:37 dod joined #mojo
19:43 dod1 joined #mojo
19:56 Janos joined #mojo
20:17 cuechan joined #mojo
20:56 cuechan joined #mojo
21:47 cuechan joined #mojo

| Channels | #mojo index | Today | | Search | Google Search | Plain-Text | summary