[Opensim-users] Load Testing OpenSim

[Mister Blue]

Could also be comparing one version of OpenSimulator against another to
check for regressions or improvements.

Testing requires setting up the system under test, creating all the test bot
accounts, loading a standard sim configuration, setting up data gathering,
firing off one or more clients to exercise the system and then collecting
and analyzing the gathered data. Each of the steps presents its own
challenge to automation.

-- mb

On Tue, Jun 22, 2010 at 10:14 AM, Dirk Krause <dirk.krause at pixelpark.com>wrote:

>  yes, one can do that.
>
>
>
> But I thought we were speaking more about a load testing a particular
> server architecture (like ReactionGrid).
>
>
>
> *Von:* opensim-users-bounces at lists.berlios.de [mailto:
> opensim-users-bounces at lists.berlios.de] *Im Auftrag von *Jor3l Boa
> *Gesendet:* Dienstag, 22. Juni 2010 19:03
>
> *An:* opensim-users at lists.berlios.de
> *Betreff:* Re: [Opensim-users] Load Testing OpenSim
>
>
>
> We also need a pre defined region with  textures, objects (physical/non)
> and scripts running (just like a normal scenario), a 'default' oar will work
> I think
>
> 2010/6/22 Dirk Krause <dirk.krause at pixelpark.com>
>
> yes, this absolutely is one way to go.
>
> we successfully experimented with
> - ec2 linux instances sporting the hippo viewer
> - streaming the hippo window output as a .flv source
> - emulating keystrokes to the hippo viewer to move the avatar
>
> all in all this makes a fully qualified client that hits the OpenSimulator
> as hard as a real client does, still it is scriptable in a limited way. and
> it would be perfectly scalable via the cloud.
>
> -- Dirk
>
>
> -----Ursprüngliche Nachricht-----
> Von: opensim-users-bounces at lists.berlios.de [mailto:
> opensim-users-bounces at lists.berlios.de] Im Auftrag von peter host
> Gesendet: Dienstag, 22. Juni 2010 12:30
> An: opensim-users at lists.berlios.de
> Betreff: Re: [Opensim-users] Load Testing OpenSim
>
>
> I haven't much dug into that, but it seems to me that a scalable cloud
> infra should be perfect for what you describe : large numbers of
> clients for a limited time, *real* different ips,...
> 1 loader script piloting 'on the fly' created ec2/whaterver client
> instances, only for the time they're needed. Even for large load
> tests, the overall cost should noy go beyond a few dollars.
>
> > Testing the load of a large number of clients requires a lot of
> infrastructure -- multiple client systems and a lot of networking. That
> makes stress testing difficult to set up and to create a portable, automated
> stress tester.
> >
> > -- mb
> >
> _______________________________________________
> Opensim-users mailing list
> Opensim-users at lists.berlios.de
> https://lists.berlios.de/mailman/listinfo/opensim-users
> _______________________________________________
> Opensim-users mailing list
> Opensim-users at lists.berlios.de
> https://lists.berlios.de/mailman/listinfo/opensim-users
>
>
>
> _______________________________________________
> Opensim-users mailing list
> Opensim-users at lists.berlios.de
> https://lists.berlios.de/mailman/listinfo/opensim-users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://opensimulator.org/pipermail/opensim-users/attachments/20100622/c416b5a5/attachment.html>