Hi Seb, I'm just wondering if you did indeed think there was a problem re. how we've configured OM looking at our stress-test results?
Thank you! Sam & doug On 1 Sep 2014, at 17:01, Sam Naef <[email protected]> wrote: > Hi Seb, > > Please see below & let me know if I can help at all > > >> On 1 Sep 2014, at 09:03, [email protected] wrote: >> >>> Can you give some more details on the test? >> of course - i'll try my best, I am not a programmer but if you know where I >> can locate more deets please let me know. >> Here is a link to the stress-test vid:https://vimeo.com/104936979 (you might >> have to wait before the video is processed) >> >> >>> I am not sure if I understand this correctly. >>> All of those CPU usage is server side only right? >> correct >>> And the server is only doing OpenMeetings, you do not run the client and >>> server on the same machine? >>> >> Correct, clients are mostly UK based, each running one OM window each. >> (However all 15 users were using the same login..) >> >>> Why would the video streaming take so much CPU power? Red5 should not do >>> any kind of transcoding. We have done experiments in the past with 100 >>> browsers connected. Obviously each browser run on its own pc/device. Not in >>> multiple tabs. Otherwise you have glitches in your stream, but he reason >>> for that is that the clients are running out of CPU, not the server. >>> >>> So this is kind of confusing, maybe you should clarify. >>> >>> Or is that an issue that is introduced since we updated to later version of >>> red5? >>> >>> Thanks, >>> Sebastian >>> On Sep 1, 2014 6:56 AM, "Maxim Solodovnik" <[email protected]> wrote: >>> >>>> you can do "external" load balancing or use OM cluster >>>> http://openmeetings.apache.org/Clustering.html >>>> >>>> >>>> On 1 September 2014 01:53, Sam Naef <[email protected]> wrote: >>>> >>>>> ah! >>>>> That could be it then :) >>>>> Ok so my results seem to be consistent it seems. >>>>> >>>>> The idea for us is to perhaps have multiple servers and do some kind of >>>>> load-balancing to direct users to a free server (so that we can scale >>>>> usage). >>>>> We're going to test, but hopefully multiple servers of 1core2GB should >>>>> hopefully work as it seems that processing power is what OM needs >>>>> >>>>> thanks >>>>> Sam & Doug >>>>> >>>>> +44 7707575060 >>>>> iSKILLu founder >>>>> Facebook >>>>> Twitter >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> On 31 Aug 2014, at 19:32, Maxim Solodovnik <[email protected]> wrote: >>>>> >>>>>> 1 room >>>>>> 8cores, 18GB RAM >>>>>> >>>>>> >>>>>> On 1 September 2014 01:26, Sam Naef <[email protected]> wrote: >>>>>> >>>>>>> yes it's version OM 3.0.3 >>>>>>> >>>>>>> When I test using multiple tabs on my browser, my computer can't hack >>>>> more >>>>>>> than about 4 tabs (tho i have fast mid 2012 macbook air) >>>>>>> >>>>>>> - were all your 20 users in the same room? >>>>>>> Should 2 cores & 4GB be working for 20 sim. users then? >>>>>>> >>>>>>> >>>>>>> >>>>>>> +44 7707575060 >>>>>>> iSKILLu founder >>>>>>> Facebook >>>>>>> Twitter >>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>>> On 31 Aug 2014, at 15:26, Maxim Solodovnik <[email protected]> >>>>> wrote: >>>>>>> >>>>>>>> What version of OM have you tested? >>>>>>>> we were able to have room with ~20video users (simulated by multiple >>>>>>>> browsers tabs) >>>>>>>> I believe 3.0.3 latest version should work better than 3.0.2 >>>>>>>> >>>>>>>> >>>>>>>> On 31 August 2014 20:35, Sam Naef <[email protected]> wrote: >>>>>>>> >>>>>>>>> Hi Maxim et all, >>>>>>>>> >>>>>>>>> We have just carried out our first largeish user test of OM :) >>>>>>>>> >>>>>>>>> >>>>>>>>> Here are our results: >>>>>>>>> We are using 2 cores, 4GB ram vps >>>>>>>>> >>>>>>>>> Users in the same room, we were able to get: >>>>>>>>> 7vids, 9 people+audio >>>>>>>>> 50-70% processers usage >>>>>>>>> mostly OK >>>>>>>>> >>>>>>>>> 9 vids, 15 people +audio >>>>>>>>> 90% of processors usage >>>>>>>>> massive glitching >>>>>>>>> >>>>>>>>> Users in different rooms: >>>>>>>>> When we compared 6 people in the same room to 3 rooms of 2 each, the >>>>>>>>> processing usage dropped by 50% (i think this is expected as less >>>>>>> bandwidth >>>>>>>>> is being sent?) >>>>>>>>> >>>>>>>>> I'm also guessing that you don't have a program that replicates a >>>>>>>>> stress-test by any chance…? >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> Kindest regards! >>>>>>>>> Sam & Doug >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>> -- >>>>>>>> WBR >>>>>>>> Maxim aka solomax >>>>>>> >>>>>>> >>>>>> >>>>>> >>>>>> -- >>>>>> WBR >>>>>> Maxim aka solomax >>>>> >>>>> >>>> >>>> >>>> -- >>>> WBR >>>> Maxim aka solomax >>>> >> >
