| *** vincentfrancoise has quit IRC | 00:00 | |
| *** vmahe has joined #openstack-watcher | 00:00 | |
| *** thorst has quit IRC | 00:25 | |
| *** hidekazu has joined #openstack-watcher | 00:29 | |
| *** vmahe has quit IRC | 01:00 | |
| *** vmahe has joined #openstack-watcher | 01:01 | |
| *** vmahe has quit IRC | 01:31 | |
| *** vmahe has joined #openstack-watcher | 01:31 | |
| *** openstackgerrit has joined #openstack-watcher | 01:40 | |
| openstackgerrit | Hidekazu Nakamura proposed openstack/watcher-specs master: Add spec for Zone migration strategy https://review.openstack.org/457834 | 01:40 |
|---|---|---|
| openstackgerrit | Merged openstack/puppet-watcher master: Add the setting log level https://review.openstack.org/457450 | 01:41 |
| *** Yumeng_ has joined #openstack-watcher | 01:58 | |
| *** Yumeng has quit IRC | 01:58 | |
| *** thorst has joined #openstack-watcher | 02:00 | |
| *** vmahe has quit IRC | 02:00 | |
| *** vmahe has joined #openstack-watcher | 02:01 | |
| *** thorst has quit IRC | 02:16 | |
| *** vmahe has quit IRC | 02:30 | |
| *** Tamayo has joined #openstack-watcher | 02:53 | |
| *** sanfern has joined #openstack-watcher | 03:02 | |
| *** hidekazu has quit IRC | 03:10 | |
| *** thorst has joined #openstack-watcher | 03:13 | |
| *** thorst has quit IRC | 03:32 | |
| *** hidekazu has joined #openstack-watcher | 04:17 | |
| *** adisky_ has joined #openstack-watcher | 04:26 | |
| *** thorst has joined #openstack-watcher | 05:29 | |
| *** zhurong has joined #openstack-watcher | 05:42 | |
| *** thorst has quit IRC | 05:48 | |
| openstackgerrit | praveenraj proposed openstack/watcher master: [WIP] Support workload fingerprint generation https://review.openstack.org/448402 | 05:52 |
| *** thorst has joined #openstack-watcher | 06:45 | |
| *** thorst has quit IRC | 06:49 | |
| *** vmahe has joined #openstack-watcher | 06:52 | |
| *** Yumeng_ has quit IRC | 07:11 | |
| *** danpawlik has joined #openstack-watcher | 07:15 | |
| *** vincentfrancoise has joined #openstack-watcher | 07:29 | |
| *** vmahe has quit IRC | 07:30 | |
| *** Yumeng has joined #openstack-watcher | 07:39 | |
| *** thorst has joined #openstack-watcher | 07:46 | |
| *** thorst has quit IRC | 07:50 | |
| *** vmahe has joined #openstack-watcher | 07:51 | |
| *** vincentfrancoise has quit IRC | 08:00 | |
| *** alexchadin has joined #openstack-watcher | 08:02 | |
| *** vincentfrancoise has joined #openstack-watcher | 08:10 | |
| *** vmahe has quit IRC | 08:10 | |
| *** vmahe has joined #openstack-watcher | 08:11 | |
| *** efoley has joined #openstack-watcher | 08:27 | |
| *** efoley_ has joined #openstack-watcher | 08:27 | |
| *** efoley_ has quit IRC | 08:27 | |
| *** alexchadin has quit IRC | 08:34 | |
| *** alexchadin has joined #openstack-watcher | 08:43 | |
| *** thorst has joined #openstack-watcher | 08:47 | |
| *** thorst has quit IRC | 08:52 | |
| *** hidekazu has left #openstack-watcher | 08:59 | |
| *** vmahe has quit IRC | 09:00 | |
| *** vmahe has joined #openstack-watcher | 09:01 | |
| openstackgerrit | praveenraj proposed openstack/watcher master: [WIP] Support workload fingerprint generation https://review.openstack.org/448402 | 09:01 |
| *** vincentfrancoise has quit IRC | 09:30 | |
| *** vmahe has quit IRC | 09:31 | |
| *** vincentfrancoise has joined #openstack-watcher | 09:31 | |
| *** thorst has joined #openstack-watcher | 09:47 | |
| *** thorst has quit IRC | 09:52 | |
| *** vincentfrancoise has quit IRC | 10:00 | |
| *** vincentfrancoise has joined #openstack-watcher | 10:28 | |
| *** alexchadin has quit IRC | 10:32 | |
| *** zhurong has quit IRC | 10:34 | |
| *** vmahe has joined #openstack-watcher | 10:42 | |
| *** thorst has joined #openstack-watcher | 10:48 | |
| *** thorst has quit IRC | 10:53 | |
| *** sanfern has quit IRC | 10:55 | |
| *** vmahe has quit IRC | 11:00 | |
| *** thorst has joined #openstack-watcher | 11:42 | |
| *** thorst_ has joined #openstack-watcher | 11:42 | |
| *** thorst has quit IRC | 11:47 | |
| *** vmahe has joined #openstack-watcher | 11:53 | |
| *** vincentfrancoise has quit IRC | 12:00 | |
| *** vincentfrancoise has joined #openstack-watcher | 12:01 | |
| *** efoley_ has joined #openstack-watcher | 12:02 | |
| *** efoley has quit IRC | 12:05 | |
| *** alexchadin has joined #openstack-watcher | 12:09 | |
| *** sanfern has joined #openstack-watcher | 12:30 | |
| *** zhurong has joined #openstack-watcher | 12:31 | |
| openstackgerrit | Merged openstack/watcher master: [Doc] messaging -> messagingv2 https://review.openstack.org/457434 | 12:40 |
| *** zhurong has quit IRC | 12:45 | |
| *** vmahe has left #openstack-watcher | 12:50 | |
| *** efoley__ has joined #openstack-watcher | 12:58 | |
| *** efoley_ has quit IRC | 13:01 | |
| *** vmahe has joined #openstack-watcher | 13:01 | |
| *** vmahe has quit IRC | 13:06 | |
| *** zhurong_phone has joined #openstack-watcher | 13:10 | |
| *** zhurong_phone has left #openstack-watcher | 13:11 | |
| openstackgerrit | Merged openstack/watcher-dashboard master: Updated from global requirements https://review.openstack.org/457737 | 13:20 |
| openstackgerrit | Merged openstack/watcher-dashboard master: Fix exception error when creating audit without audit template https://review.openstack.org/455679 | 13:20 |
| *** alexchadin has quit IRC | 13:21 | |
| *** vincentfrancoise has quit IRC | 13:30 | |
| *** vincentfrancoise has joined #openstack-watcher | 13:31 | |
| *** alexchadin has joined #openstack-watcher | 13:53 | |
| alexchadin | hi, we will start in 5 minutes on #openstack-meeting-4 | 13:54 |
| sballe_ | morning | 13:57 |
| alexchadin | hi sballe_ | 13:58 |
| sballe_ | hi :-) alexchadin | 13:58 |
| *** vincentfrancoise has quit IRC | 14:00 | |
| *** vincentfrancoise has joined #openstack-watcher | 14:01 | |
| alexchadin | Yumeng: adisky_ chrisspencer vincentfrancoise sanfern we have meeting on #openstack-meeting-4 channel now | 14:04 |
| alexchadin | hvprash: ^ | 14:07 |
| *** pshedimb_ has joined #openstack-watcher | 14:16 | |
| *** vincentfrancoise has quit IRC | 14:30 | |
| *** vincentfrancoise has joined #openstack-watcher | 14:31 | |
| *** sanfern has quit IRC | 14:43 | |
| *** mkarsten has joined #openstack-watcher | 14:57 | |
| *** vincentfrancoise has quit IRC | 15:00 | |
| *** vincentfrancoise has joined #openstack-watcher | 15:08 | |
| mkarsten | Hello. aspiers recommended that I join this channel. I am an academic and interested in understanding the state of the art in cloud rebalancing. I am somewhat familiar with OpenStack nuts and bolts, but certainly no expert. I do know a fair bit about resource management and performance in general. I am trying to assess the viability of dynamic rebalancing, given the overhead and latency of migrations in comparison to the frequency of | 15:09 |
| mkarsten | workload changes. For starters, the Watcher web page speaks about providing "out-of-box optimization routines for immediate value-add", so I was wondering whether these are documented somewhere and/or if there is other literature out there describing such optimization routines? | 15:09 |
| aspiers | welcome mkarsten :-) | 15:10 |
| aspiers | I'd suggest staying connected as long as possible - getting a reply sometimes takes hours, depending on the time of day and how busy people are with other stuff.. | 15:11 |
| alexchadin | good day, mkarsten | 15:12 |
| alexchadin | mkarsten: first of all, you are welcome here | 15:12 |
| aspiers | or sometimes you only have to wait a few seconds ;-) | 15:13 |
| alexchadin | aspiers: +2 | 15:13 |
| alexchadin | mkarsten: if I understand right, you'd like to get deeper into Watcher workflow process | 15:13 |
| alexchadin | mkarsten: start from viewing our demo: https://www.youtube.com/watch?v=FJttohkf0sM | 15:16 |
| alexchadin | mkarsten: then I would suggest you to learn System Architecture here: https://docs.openstack.org/developer/watcher/architecture.html | 15:19 |
| mkarsten | The workflow process is certainly part of it, but my first objective is getting a back-of-the envelope sense of: | 15:20 |
| mkarsten | - frequency and amplitude of workload changes | 15:20 |
| mkarsten | - overhead/latency of rebalancing actions (VM migrations?, other?) | 15:20 |
| alexchadin | mkarsten: the best way to get more familiar with something new is try to install it, so you can install devstack and follow Getting Started section: https://docs.openstack.org/developer/watcher/#getting-started | 15:20 |
| mkarsten | In other words, assume for a moment that there is a infinitely fast/good algorithm and we have a perfect system to execute the outcome of the algorithm. | 15:20 |
| alexchadin | mkarsten: about frequency: we rely on some data sources (Ceilometer+Gnocchi, Monasca) which gather metrics. | 15:23 |
| mkarsten | I just want to size up the opportunity, having seen many a feedback-based resource allocation scheme not gaining much traction in the real world (for example, 20 years of "QoS routing" research for the Internet - and there's not much to show for). | 15:23 |
| *** mkarsten has left #openstack-watcher | 15:23 | |
| *** mkarsten has joined #openstack-watcher | 15:23 | |
| mkarsten | What are your assumptions about the scale of the system and the workload? What kind of demand swings do you expect? How much time do you budget for a VM migration? Does anyone have some (estimated) numbers? | 15:25 |
| vincentfrancoise | mkarsten: hi | 15:26 |
| *** alexchadin has quit IRC | 15:27 | |
| *** vincentfrancoise has quit IRC | 15:31 | |
| *** vincentfrancoise has joined #openstack-watcher | 15:31 | |
| vincentfrancoise | mkarsten: the case I was the most confronted with is the live migration and we usually consider its cost to be negligible because it is actually absorbed by the syncing of NFS underneath it | 15:32 |
| vincentfrancoise | mkarsten: when we say we have out-of-the-box routines, what can be easily shown is the before/after effect | 15:34 |
| vincentfrancoise | mkarsten: but I agree that we (my team) didn't spend much time on the interference cause by these re-organization actions (e.g. migration) | 15:35 |
| mkarsten | vincentfrancoise: Thanks, I see. Do you assume any latency at all? For example, after a live migration, a VM might also experience a ramp-up phase to get back up to speed? | 15:37 |
| mkarsten | vincentfrancoise: Is there any documentation about these routines? Or do I have to go to the source (code)? | 15:38 |
| vincentfrancoise | mkarsten: hum, I'll try to answer to the best of my knowledge though I am not the best qualified to answer that | 15:40 |
| vincentfrancoise | mkarsten: the ramp-up phase you describe is something we do not model because we reduced the scope of our tests to live-migrations which had very little impact with this regard, or at least not too bad unless the VM is actually under a high load | 15:42 |
| vincentfrancoise | mkarsten: at least not in the context of Watcher | 15:42 |
| vincentfrancoise | mkarsten: the problem with modeling this is that you need a testing protocol which simulates this workload to detect any perturbation and that's hard to do | 15:43 |
| mkarsten | Right, that's why I want to start with looking at algorithms that people have in mind and simply plug in cost estimates to assess their viability under certain circumstances. | 15:45 |
| vincentfrancoise | mkarsten: again, my team was mainly focused on producing a good before/after affect, and we didn't invest much time investigating the performance impact during the state transition | 15:45 |
| vincentfrancoise | mkarsten: the algorithms (we call then strategies) are lightly described in the online doc | 15:46 |
| vincentfrancoise | mkarsten: https://docs.openstack.org/developer/watcher/dev/plugins.html#strategies | 15:47 |
| vincentfrancoise | mkarsten: for the rest you can dig in their associated spec (sometimes the URL is referenced in the doc I gave you) | 15:47 |
| mkarsten | Great, thanks, that's a good starting point! I'll take a look and will get back here with more questions... | 15:48 |
| vincentfrancoise | mkarsten: for the rest you have to read the code | 15:48 |
| vincentfrancoise | mkarsten: but any algorithm question starts to get outside of my "scope" so I may not be able to answer | 15:48 |
| openstackgerrit | praveenraj proposed openstack/watcher-specs master: Support visualizing workload fingerprints https://review.openstack.org/448394 | 15:48 |
| *** sanfern has joined #openstack-watcher | 15:53 | |
| *** wootehfoot has joined #openstack-watcher | 16:26 | |
| *** vincentfrancoise has quit IRC | 16:32 | |
| *** ss4 has joined #openstack-watcher | 16:37 | |
| *** wootehfoot has quit IRC | 16:40 | |
| *** harlowja_ has quit IRC | 17:13 | |
| *** harlowja has joined #openstack-watcher | 17:15 | |
| *** harlowja has quit IRC | 17:17 | |
| *** harlowja has joined #openstack-watcher | 17:17 | |
| *** efoley__ has quit IRC | 17:18 | |
| *** mkarsten has left #openstack-watcher | 17:29 | |
| *** pshedimb_ has quit IRC | 17:48 | |
| *** mkarsten has joined #openstack-watcher | 17:52 | |
| *** adisky_ has quit IRC | 18:19 | |
| *** mkarsten has left #openstack-watcher | 18:36 | |
| *** mkarsten has joined #openstack-watcher | 18:38 | |
| *** pshedimb_ has joined #openstack-watcher | 18:46 | |
| *** pshedimb_ has quit IRC | 18:53 | |
| *** wootehfoot has joined #openstack-watcher | 19:02 | |
| *** ss4 has quit IRC | 19:06 | |
| *** pshedimb_ has joined #openstack-watcher | 19:12 | |
| *** wootehfoot has quit IRC | 19:27 | |
| *** wootehfoot has joined #openstack-watcher | 19:32 | |
| *** wootehfoot has quit IRC | 19:53 | |
| *** wootehfoot has joined #openstack-watcher | 19:55 | |
| *** wootehfoot has quit IRC | 19:59 | |
| *** wootehfoot has joined #openstack-watcher | 20:05 | |
| *** mkarsten has left #openstack-watcher | 20:13 | |
| *** thorst_ has quit IRC | 21:10 | |
| *** thorst has joined #openstack-watcher | 21:35 | |
| *** thorst has quit IRC | 21:40 | |
| *** wootehfoot has quit IRC | 22:33 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!