| *** devlaps has joined #openstack-swift | 00:23 | |
| *** devlaps has quit IRC | 00:27 | |
| *** _bluev has quit IRC | 00:39 | |
| *** _bluev has joined #openstack-swift | 00:39 | |
| *** _bluev1 has joined #openstack-swift | 00:51 | |
| *** _bluev has quit IRC | 00:52 | |
| *** charlesw has joined #openstack-swift | 00:53 | |
| portante | swifterdarrell: perhaps, I might be, the work to associate a port with a drive sounds interesting, though you still have to deal with one of the object servers sucking up accepts for incoming requests when it gets backed up | 01:27 |
|---|---|---|
| portante | Ignore me if this does not make sense, but I think we still have to deal with how often an object server (or really any server in the system) thinks it can accept another request | 01:29 |
| portante | If there are no existing requests, accept, if there are existing requests and they are progressing fine (hand way on that definition) accept, but if there are existing request in flight but they are stalled, adding another for the same disk is prolly not a good idea | 01:31 |
| *** zynisch_o7 has joined #openstack-swift | 01:42 | |
| *** _bluev1 has quit IRC | 01:43 | |
| portante | swifterdarrell: with regards to RHEL features, I'd be willing to help lobby to get them back-ported to RHEL | 01:45 |
| portante | that is what we often do on the performance team, document the performance gain to help the lobbying process | 01:45 |
| *** zynisch_o7 has quit IRC | 01:47 | |
| swifterdarrell | portante: *nod* clayg was telling me about how the go server limits outstanding reqs per disk, and getting some kind of backpressure to the proxy when the object-server (once it's only dealing w/a single disk) has some threshold of requests in-flight would probably also help | 02:00 |
| portante | yes, sounds promising | 02:01 |
| *** bkopilov has quit IRC | 02:26 | |
| openstackgerrit | Michael Barton proposed openstack/swift: go: ring self-reloading https://review.openstack.org/187030 | 02:37 |
| openstackgerrit | Michael Barton proposed openstack/swift: go: ring self-reloading https://review.openstack.org/187030 | 02:40 |
| *** zynisch_o7 has joined #openstack-swift | 03:31 | |
| *** zynisch_o7 has quit IRC | 03:35 | |
| *** haomaiwang has quit IRC | 03:46 | |
| *** charlesw has quit IRC | 04:00 | |
| *** bkopilov has joined #openstack-swift | 04:12 | |
| *** zynisch_o7 has joined #openstack-swift | 04:31 | |
| *** zynisch_o7 has quit IRC | 04:36 | |
| *** SkyRocknRoll has joined #openstack-swift | 05:56 | |
| *** jamielennox is now known as jamielennox|away | 06:17 | |
| *** haomaiwa_ has joined #openstack-swift | 06:18 | |
| *** zynisch_o7 has joined #openstack-swift | 06:20 | |
| *** zynisch_o7 has quit IRC | 06:25 | |
| *** SkyRocknRoll has quit IRC | 06:32 | |
| *** SkyRocknRoll has joined #openstack-swift | 06:54 | |
| *** gsilvis has quit IRC | 07:07 | |
| *** gsilvis has joined #openstack-swift | 07:09 | |
| *** gvernik has joined #openstack-swift | 07:57 | |
| *** dimasot has joined #openstack-swift | 07:58 | |
| gvernik | hi dimasot | 07:59 |
| *** gvernik has left #openstack-swift | 08:03 | |
| *** haomaiwa_ has quit IRC | 08:09 | |
| *** zynisch_o7 has joined #openstack-swift | 08:09 | |
| *** zynisch_o7 has quit IRC | 08:13 | |
| *** Fiatx has joined #openstack-swift | 09:02 | |
| *** Fiatx has left #openstack-swift | 09:03 | |
| *** silor has joined #openstack-swift | 09:06 | |
| *** cdelatte has quit IRC | 09:25 | |
| *** geaaru has joined #openstack-swift | 09:44 | |
| *** zynisch_o7 has joined #openstack-swift | 09:58 | |
| *** zynisch_o7 has quit IRC | 10:02 | |
| *** cdelatte has joined #openstack-swift | 10:08 | |
| *** delattec has joined #openstack-swift | 10:08 | |
| *** delatte has joined #openstack-swift | 10:23 | |
| *** delattec has quit IRC | 10:23 | |
| *** cdelatte has quit IRC | 10:24 | |
| *** zynisch_o7 has joined #openstack-swift | 11:46 | |
| *** zynisch_o7 has quit IRC | 11:51 | |
| *** jrichli has joined #openstack-swift | 12:32 | |
| *** bkopilov has quit IRC | 12:37 | |
| *** jamielennox|away is now known as jamielennox | 12:40 | |
| *** zynisch_o7 has joined #openstack-swift | 13:02 | |
| *** proteusguy has quit IRC | 13:06 | |
| *** zynisch_o7 has quit IRC | 13:07 | |
| *** cdelatte has joined #openstack-swift | 13:07 | |
| *** delatte has quit IRC | 13:08 | |
| *** delattec has joined #openstack-swift | 13:10 | |
| *** cdelatte has quit IRC | 13:11 | |
| *** delattec has quit IRC | 13:17 | |
| *** jamielennox is now known as jamielennox|away | 13:19 | |
| *** proteusguy has joined #openstack-swift | 13:22 | |
| *** proteusguy has quit IRC | 13:30 | |
| *** proteusguy has joined #openstack-swift | 13:42 | |
| *** proteusguy has quit IRC | 13:43 | |
| *** haomaiwang has joined #openstack-swift | 13:43 | |
| *** proteusguy has joined #openstack-swift | 13:43 | |
| *** charlesw has joined #openstack-swift | 14:10 | |
| *** SkyRocknRoll has quit IRC | 14:23 | |
| *** bkopilov has joined #openstack-swift | 14:31 | |
| *** archers has joined #openstack-swift | 14:44 | |
| *** zynisch_o7 has joined #openstack-swift | 14:48 | |
| *** SkyRocknRoll has joined #openstack-swift | 14:49 | |
| *** zynisch_o7 has quit IRC | 14:53 | |
| *** cdelatte has joined #openstack-swift | 15:25 | |
| *** zynisch_o7 has joined #openstack-swift | 15:49 | |
| *** charlesw has quit IRC | 15:51 | |
| *** jrichli has quit IRC | 15:52 | |
| *** zynisch_o7 has quit IRC | 15:53 | |
| *** ajiang has quit IRC | 16:02 | |
| *** bhakta has quit IRC | 16:02 | |
| *** ajiang has joined #openstack-swift | 16:08 | |
| *** bhakta has joined #openstack-swift | 16:09 | |
| *** archers has quit IRC | 16:28 | |
| *** delattec has joined #openstack-swift | 16:48 | |
| *** cdelatte has quit IRC | 16:49 | |
| *** delattec has quit IRC | 16:53 | |
| *** devlaps has joined #openstack-swift | 16:53 | |
| *** cdelatte has joined #openstack-swift | 16:59 | |
| *** delattec has joined #openstack-swift | 16:59 | |
| *** tab___ has joined #openstack-swift | 17:02 | |
| *** devlaps has quit IRC | 17:09 | |
| *** charlesw has joined #openstack-swift | 17:20 | |
| *** zynisch_o7 has joined #openstack-swift | 17:37 | |
| *** zynisch_o7 has quit IRC | 17:42 | |
| swifterdarrell | clayg: notmyname: portante: zaitcev: prelim results (2048 clients, 64KiB GET-only across 10M objects): https://gist.github.com/dbishop/1d14755fedc86a161718#file-tabular_results-md | 18:29 |
| swifterdarrell | clayg: notmyname: portante: zaitcev: I'll be exploring further, but I think the gain for servers_per_port >=2 are clear. The sweet spot for my setup seems to be 3 or 4. | 18:30 |
| swifterdarrell | clayg: notmyname: portante: zaitcev: (that gist has more details of my setup, configs, etc.) | 18:31 |
| *** silor1 has joined #openstack-swift | 18:53 | |
| *** silor has quit IRC | 18:55 | |
| *** delattec has quit IRC | 18:57 | |
| *** delattec has joined #openstack-swift | 18:57 | |
| *** charlesw has quit IRC | 19:19 | |
| *** dimasot has quit IRC | 19:21 | |
| *** SkyRocknRoll has quit IRC | 19:25 | |
| *** zynisch_o7 has joined #openstack-swift | 19:26 | |
| *** zynisch_o7 has quit IRC | 19:31 | |
| ctennis | based on the std dev being lower and more close to the avg I'd say that at the very least the workflow is much more deterministic | 19:42 |
| *** proteusguy has quit IRC | 20:06 | |
| *** proteusguy has joined #openstack-swift | 20:12 | |
| *** _bluev has joined #openstack-swift | 20:50 | |
| *** cdelatte has quit IRC | 20:50 | |
| *** tab___ has quit IRC | 20:51 | |
| *** delattec has quit IRC | 20:51 | |
| *** 16WAA4FX8 has joined #openstack-swift | 20:56 | |
| *** 17WABZSPI has joined #openstack-swift | 20:56 | |
| *** zynisch_o7 has joined #openstack-swift | 21:15 | |
| *** torgomatic has quit IRC | 21:17 | |
| *** theanalyst has quit IRC | 21:17 | |
| *** torgomatic has joined #openstack-swift | 21:17 | |
| *** ChanServ sets mode: +v torgomatic | 21:17 | |
| *** thurloat_isgone has quit IRC | 21:18 | |
| *** thurloat_isgone has joined #openstack-swift | 21:19 | |
| *** thurloat_isgone is now known as thurloat | 21:19 | |
| *** zynisch_o7 has quit IRC | 21:19 | |
| *** theanalyst has joined #openstack-swift | 21:20 | |
| *** _bluev has quit IRC | 21:22 | |
| *** evanjfraser has quit IRC | 21:26 | |
| *** Guest84 has joined #openstack-swift | 21:33 | |
| *** Guest84_ has joined #openstack-swift | 21:50 | |
| *** Guest84_ has joined #openstack-swift | 21:50 | |
| *** Guest84_ has quit IRC | 21:53 | |
| *** Guest84 has quit IRC | 21:53 | |
| *** bi_fa_fu has joined #openstack-swift | 21:54 | |
| *** geaaru has quit IRC | 22:07 | |
| mattoliverau | Morning | 22:13 |
| portante | swifterdarrell: nice, 50% speedup with 3 servers_per_port, how many disks? And what is the maximum disks we can handle with ports-per-disk for a server? | 22:13 |
| *** jamielennox|away is now known as jamielennox | 22:17 | |
| portante | swifterdarrell: and have you considered seeing how close to a 10Gb line speed you can sustain varying the number of disks? | 22:31 |
| *** proteusguy has quit IRC | 22:41 | |
| portante | swifterdarrell: found the disk counts in the gist, so what about 128 and 256 K request sizes? I'd be curious how this fares when the configured object server chunk size is smaller than the request. | 22:45 |
| portante | We also want to make sure that small requests to a given disk don't starve out larger requests, as there is a tendency for that to happen with our model | 22:46 |
| *** proteusguy has joined #openstack-swift | 22:54 | |
| *** km has joined #openstack-swift | 22:54 | |
| *** _bluev has joined #openstack-swift | 22:56 | |
| *** silor1 has quit IRC | 22:56 | |
| *** zynisch_o7 has joined #openstack-swift | 23:04 | |
| *** zynisch_o7 has quit IRC | 23:08 | |
| *** kei_yama has joined #openstack-swift | 23:12 | |
| *** chlong has joined #openstack-swift | 23:44 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!