Hi Dan, I am using the latest trunk and I tried to allocate as much RAM as I could to red5. You are right, playing the streams from the flash client might really max out the CPU cycles. I try and see what happens.
On 3/15/07, Dan Rossi <[EMAIL PROTECTED]> wrote:
Wow interesting, ill check this out. This used to do this, are you using the latest trunk and have allocated pretty much all your ram to red5 :D Yes it seems the server is quite cpu/ram intensive im having to order 4GB of ram on the slave machines im going to be migrating red5 to because its currently on a standalone server right now. joseph wamicha wrote: > Hi, > > Preliminary results of the load test show red5 running out of memory > (errors > can be seen in URL below) when it publishes several live streams. > > The current load test sources version can be found here: > http://www.korandu.com/red5loadtests/red5loadtests.html > > I will continue to try and improve and refactor the load test code. > > Memory runs out before initiating play from the flash end. It seems > FLVReader is the one that seems to causes the out of memory errors. > > I am using a 1.8GHz celeron, 756MB laptop. The more streams I > initialize the > more in all honesty I hear my machine throttle (same like when you really > push your car engine hard); you can actually even "hear" the load test > :-) > > Tabulated results: > Number of live streams Out of Memory > - - - - - - - - - - - - - - - - - - - - - - - - - - - - - > 10 No > 20 No > 30 No > 40 No > 50 No > 60 No > 70 No > 80 No > 90 No > > 100 No > 110 No > 120 No > 130 Yes > 200 Yes > > Next up would be testing how red5 is able to handle subscription to > several > live streams or even several subscriptions to a single live stream (ie > how > many clients it can comfortably support). > > The results are quite encouraging though because 120 simultaneous live > streams (though simultaneous subscription is not yet tested) is not bad > IMHO? > > Who might have a more powerful machine available to do load tests? > > On 3/14/07, joseph wamicha <[EMAIL PROTECTED]> wrote: >> >> Hi Chris, >> >> I'm trying out the tests right now. I'll let you know how it goes. >> >> Hello Thijs, how may I help? >> >> On 3/14/07, Chris Allen < [EMAIL PROTECTED]> wrote: >> > >> > Hi Joseph, >> > >> > You may want to work with Thijs on this one, as he's been heading up >> > testing for Red5. He may already have something like this in place. >> > >> > At any rate, I would love to hear how your test goes. >> > >> > -Chris >> > >> > On 3/13/07, joseph wamicha <[EMAIL PROTECTED]> wrote: >> > > Hello, >> > > >> > > I am sure this would be a good stress test on the number of >> > simultaneous >> > > streams that red5 can handle: >> > > >> > > Create an array of serverStreams which are live streams, >> > > then, >> > > Initialize all those live streams by looping through array. >> > > So if you want like 100 live streams done by red5 do: >> > > >> > > private IServerStream[] serverStream; >> > > >> > > for(int i=0; i<100; i++) >> > > { >> > > serverStream[i] = StreamUtils.createServerStream(appScope, >> > "live0"); >> > > SimplePlayItem item = new SimplePlayItem(); >> > > item.setName("on2_flash8_w_audio"); >> > > serverStream[i].addItem(item); >> > > item = new SimplePlayItem(); >> > > item.setName("on2_flash8_w_audio"); >> > > serverStream[i].addItem(item); >> > > serverStream[i].start(); >> > > } >> > > >> > > I'm fairly certain this should work. It will produce 100 live >> streams >> > by >> > > red5. You could scale it to as many as you want, even 1000, 10000 >> > etc... >> > > >> > > Now on the client end do a similar array to subscribe to the stream >> > even if >> > > you don't play it just for the sake of load test. >> > > >> > > If someone could write a mock client flash end for subscribing to >> > these >> > > streams by declaring an array of Netstreams, it would form a very >> good >> > load >> > > test. We could then toggle it higher bit by bit and see how red5 >> > handles >> > > under load and what code breaks. I'll write this load test code >> today >> > night >> > > but I hope someone else can beat me to it. >> > > >> > > >> > > On 3/14/07, Ruben Waitz <[EMAIL PROTECTED]> wrote: >> > > > Hello, >> > > > >> > > > Just wondering if somebody has FMS2 and Red5 both installed on a >> > single >> > > > system. >> > > > I'm curious about the server load and performance differences in >> > both >> > > > situations (benchmarking). Maybe a dedicated "stresstest" SWF can >> > act as a >> > > > client. >> > > > >> > > > I don't know whether such a client is already developed by someone >> > but I >> > > > think benchmarking results are quite interesting for everyone. >> > > > >> > > > Ruben >> > > > >> > > > ------------------------------------ >> > > > www.red5tutorials.net: Tutorials - How tos - FAQ >> > > > >> > > > >> > > > _______________________________________________ >> > > > Red5 mailing list >> > > > [email protected] >> > > > http://osflash.org/mailman/listinfo/red5_osflash.org >> > > > >> > > >> > > >> > > >> > > -- >> > > C is forever. >> > > _______________________________________________ >> > > Red5 mailing list >> > > [email protected] >> > > http://osflash.org/mailman/listinfo/red5_osflash.org >> > > >> > > >> > >> > _______________________________________________ >> > Red5 mailing list >> > [email protected] >> > http://osflash.org/mailman/listinfo/red5_osflash.org >> > >> >> >> >> -- >> C is forever. > > > > > ------------------------------------------------------------------------ > > _______________________________________________ > Red5 mailing list > [email protected] > http://osflash.org/mailman/listinfo/red5_osflash.org > _______________________________________________ Red5 mailing list [email protected] http://osflash.org/mailman/listinfo/red5_osflash.org
-- C is forever.
_______________________________________________ Red5 mailing list [email protected] http://osflash.org/mailman/listinfo/red5_osflash.org
