[Red5] Loadtest again

Thu Feb 28 04:50:20 PST 2008

Sergey,

Interesting. Thanks for the info.

Question. The items in Red, do you a log of what caused
the errors logged?

Can your test be adapted to videos on demand?

What are you using to graph the connections?

Thanks,

Lenny

On 2/28/08, Mamontov Sergey V. <moonlight at ua.elro.com> wrote:
>
> Hello,
>
> today we was interested Red5 again, but not remotely, because of
> problems with bandwidth. And I want share with it, maybe it useless
> but we found it interested, and believe what everything is OK with
> streaming ability of Red5 :)
>
> The point of test was to determine maximum available parallel
> streaming connections.
> During test was used my computer (Intel Core 2 Duo 2.2GHz, 2Gb of
> memory, Red5 0.6.3 rev. 2670, SATAII drive). We write simple
> application witch log every second count of connections)
> At client side we make application, witch start 20 connections and by
> 1 stream on each, next connections initiates only after previous
> successfully connected. In case of "disconnect", "reject", "track
> finished" client swf pull specified url and write down client log.
> After connection, client receives list of available tracks and pick 20
> of random.
>
> We asked 8 of our co-workers (in our local network) open this client and
> monitor them, all
> clients successfully established all connections, and all streams are
> start playing (8*20 = 160 MP3 streams MP3). As I saw, Dan write what
> using of IP instead of computers name are more effective, so we did
> this test with my server IP.
>
> Result of test in attachment.
> Blue line is number of connections witch currently connected to
> server.
> Red line is summary count of errors at this moment of time.
> This two read pick was noticed, when server CPU load was 100% and hi
> drop some connections, client in 5 second reconnect them again.
>
>
> Looks like not bad as I think :)
>
> Thanks,
>
> Regards,
>
> Sergey Mamontov
>
> > Sergey,
> >
> >  OK.
> >
> >  Thanks for your response.
> >
> >  Lenny
>
> >  On Tue, Feb 26, 2008 at 8:22 AM, Mamontov Sergey V. <
> moonlight at ua.elro.com> wrote:
> >
> > Lenny,
>
> > maybe we cam help, but we have some problems with Internet, Pilosoft
> > un-respect!
> > If it will be need, we maybe can give our server, but it look's like
> > we should speak with CEO.
>
> > Regards,
>
> > Sergey Mamontov
> >
> >
> >
>
>
> >> Hello Sergey,
> >>
> >>  Hang on  buddy.
> >>
> >>  I and hopefully with others am attempting to organize a testing effort
> now.
> >>
> >>  Maybe we can pull one together that will benefit everyone.
> >>
> >>  Thanks for your quick response.
> >>
> >>  Lenny
>
> >>  On Tue, Feb 26, 2008 at 7:28 AM, Mamontov Sergey V. <
> moonlight at ua.elro.com> wrote:
> >>
> >> Hello Lanny,
>
> >> It will be great, but because of problems with Pilosoft, we have small
> >> bandwidth :( But if only check individual connection and problem with
> >> "Disconnected due large inactivity" maybe it will be good idea :) But
> >> I will need to consult with my boss :)
>
> >> If this problem occurs only when simulate large count of connection
> >> from one machine, my question is no need in answer :)
> >>
> >> And also, if it possible, can someone recommend software, for
> >> registration loadtest results like: CPU pick load, memory load,
> >> filesystem usage and so on?
>
> >> Regards,
>
> >> Sergey Mamontov
> >>
> >>
> >>
>
> >>> Sergey,
> >>>
> >>>  I will be glad to be a tester today if you need someone
> >>>  to help reach the 250 test limit.
> >>>
> >>>  I too have seen when connections are made individually,
> >>>  I don't see any immediate problems.
> >>>
> >>>  If you can set a time for people to hit your test server I will
> >>>  set aside time to test.
> >>>
> >>>  Regards,
> >>>
> >>>  Lenny
> >>>
> >>>
>
> >>>
> >>>  On 2/26/08, Mamontov Sergey V. <moonlight at ua.elro.com> wrote:
> >>> Hello,
>
> >>> today due project needs we made loadtest.
> >>> Machine was:
> >>> - under Windows Server 2003
> >>>  - Red5 0.6.3 () under Jetty
> >>> - 2.33Ghz Dual Core Xeon x 2
> >>> - 4Gb DDR3
> >>> - 14Tb RAID 5EE
> >>> - Internet (some problems on Pilosoft) about 1096 kbps (it really bad
> >>> today :( )
> >>> - Streaming files are MP3 192 VBR (stored locally, and picked up
> >>>  randomly from 100 available for test)
> >>> - Server physically located USA (Bethany Beach in Delaware) tester in
> Ukraine
> >>
> >>> We used our own loadTest application, witch make new stream only when
> >>> previous created, one by one. Our mission is was to understand is Red5
> >>>  able to hold a lot of streams.
> >>> We noticed when start 10 NetConnection and by 1 stream on each at same
> >>> machine, the most of them loose connection with server (server
> >>> error.log show "Disconnected due long inactivity")
> >>>  At same time, if we try to launch by 1 NetConnection and 1 stream on
> >>> it, and make this operation on 10 different computers they all work
> >>> great, none of them loose connection. We only have some problems,
> >>> buffer refill very often (maybe caused by problems in Pilosoft)
> >>>
> >>> I don't know, maybe problems with disconnect caused by Flash player
> >>> itself? In our test it looks like true.
>
> >>> If someone can handle in this problem, we will try make more test with
> >>> about 250 connection and streams.
> >>>
> >>> Regards,
>
> >>> Sergey Mamontov
> >>
>
> >>> _______________________________________________
> >>> Red5 mailing list
> >>> Red5 at osflash.org
> >>> http://osflash.org/mailman/listinfo/red5_osflash.org
> >>>
>
> >>>
>
>
> >>
> >> --
> >> С уважением,
> >>  Mamontov                          mailto:moonlight at ua.elro.com
>
>
> >> _______________________________________________
> >>
> >>
> >>
> >> Red5 mailing list
> >> Red5 at osflash.org
> >> http://osflash.org/mailman/listinfo/red5_osflash.org
> >>
>
>
> >>
>
>
>
> > _______________________________________________
> > Red5 mailing list
> > Red5 at osflash.org
> > http://osflash.org/mailman/listinfo/red5_osflash.org
> >
>
> _______________________________________________
> Red5 mailing list
> Red5 at osflash.org
> http://osflash.org/mailman/listinfo/red5_osflash.org
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://osflash.org/pipermail/red5_osflash.org/attachments/20080228/efbad04c/attachment-0001.html 