Hi Kabelo, The tests you refer to [1] are actually quite old. Since then, we’ve updated the performance tests to include many more of our REST endpoints, including our POST endpoints. The different balance and rate of requests caused by that changed the end results, and we do have some much more recent test results that you can look at (e.g. a recent test around message body searching [2]). These tests also include the Munin results for most of the boxes.
> The Tsung tests do not include etherpad as Tsung doesn't have great > support for websockets (yet). Simon has worked on this with the Tsung > community for some new features, but I still think a useful etherpad > protocol test using Tsung would be difficult to get accurate. When we introduced Etherpad, Simon did run a suite of tests that gave us an idea of how many concurrent users we expect to be able to support per Etherpad node. Simon will need to jump in to provide the exact number we came up with, but I remember that it was definitely lower than the number of concurrent users our app nodes themselves can sustain. That is not a big problem in itself though, as we can scale them out horizontally and they don’t tend to be very expensive to run. Depending on the usage profile you’re seeing, it might just mean that you have to scale them out more than some of the other nodes. [1] http://www.slideshare.net/nicolaasmatthijs/apereo-oae-bootcamp [2] http://37.153.96.172/messageSearch/master/20140108-2059/report.html Hope that helps, Nicolaas On 27 Jan 2014, at 09:12, Branden Visser <mrvis...@gmail.com> wrote: > Hi Kabelo, > >> Regarding the setup in the slides, Etherpad and RabbitMQ aren't listed. Were >> they sharing a VM with one of the other components? If so which? >> Trying to put together a setup that can repeat the results achieved in the >> Tsung test. >> > > RabbitMQ wasn't on there likely just because the Munin monitoring was > not enabled for that particular test. In general RabbitMQ is very > under-utilized in OAE, even on the smallest VM you can find from a > public cloud provider. As all other components, during our performance > tests RabbitMQ runs on a dedicated machine. > >> Side note. >> Did the Tsung test simulate Etherpad document editing? Did a stress test of >> Etherpad using [1] and it crashed because of stack size limit. The fix for >> this is increasing stack size according to a Github issue entry. Just >> curious about whether Tsung test included Etherpad. >> > > The Tsung tests do not include etherpad as Tsung doesn't have great > support for websockets (yet). Simon has worked on this with the Tsung > community for some new features, but I still think a useful etherpad > protocol test using Tsung would be difficult to get accurate. > > Thanks for the tip on increasing stack size! > > Hope that helps, > Branden > >> [1] https://bitbucket.org/rbraakman/etherpad-stresstest >> >> Thanks. >> >> >> On 24 January 2014 15:09, Branden Visser <mrvis...@gmail.com> wrote: >>> >>> Hi Kabelo, >>> >>> Yes information for CPU usage, load memory can be found for all nodes >>> if you install the munin-node service on the nodes in your cluster and >>> specify in the Tsung XML to monitor the munin stats of that node [1]. >>> When you generate the final report, it will add an additional >>> monitoring section to the report that gives you synchronized >>> performance stats with your test. >>> >>> For elasticsearch size you could query this from the elasticsearch >>> stats API during or after the test has run using the stats API [2] [3] >>> that elasticsearch provides. >>> >>> For Cassandra database, you can get more detailed information using >>> the nodetool [4] utility or a more advanced tool like DataStax >>> OpsCenter Community Edition. >>> >>> Hope that helps, >>> Branden >>> >>> [1] http://tsung.erlang-projects.org/user_manual/conf-monitoring.html >>> [2] >>> http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/cluster-nodes-stats.html >>> [3] https://gist.github.com/mrvisser/aafaa6a47366755aad31 >>> [4] https://gist.github.com/mrvisser/8596885 >>> >>> >>> >>> On Thu, Jan 23, 2014 at 7:13 AM, Kabelo Letsoalo >>> <kab...@opencollab.co.za> wrote: >>>> Hi guys, >>>> >>>> Regarding the Tsung tests ran on this setup [1] slide 84, are there >>>> system >>>> stats like CPU, RAM usage, sizes of ElasticSearch index and Cassandra DB >>>> after completion of test? >>>> >>>> [1] http://www.slideshare.net/nicolaasmatthijs/apereo-oae-bootcamp >>>> >>>> Thanks. >>>> >>>> ________________________________ >>>> See OpenCollab email disclaimer at >>>> http://www.opencollab.co.za/email-disclaimer >>>> _______________________________________________ >>>> oae-dev mailing list >>>> oae-dev@collab.sakaiproject.org >>>> http://collab.sakaiproject.org/mailman/listinfo/oae-dev >>>> >> >> >> >> >> -- >> Kabelo Letsoalo >> Junior Developer >> opencollab >> Tel: +27 21 970 4000 | Fax: +27 21 914 3098 >> Web: www.opencollab.co.za >> >> >> ________________________________ >> See OpenCollab email disclaimer at >> http://www.opencollab.co.za/email-disclaimer > _______________________________________________ > oae-dev mailing list > oae-dev@collab.sakaiproject.org > http://collab.sakaiproject.org/mailman/listinfo/oae-dev _______________________________________________ oae-dev mailing list oae-dev@collab.sakaiproject.org http://collab.sakaiproject.org/mailman/listinfo/oae-dev
