[jdev] Automated XMPP Server Benchmarks

[Michael Weibel]

Hi Steffen,

Perfect. I’ve actually done exactly the same.. 
Funny :D that’s probably why you were the first stargazer on the repository? ;-)


Although not on EC2 but internally on docker and vmware. But I have a lot of tsung load tests and also a puppet version of tigase (SPEC file etc). 
I will contribute as soon as I can with the tigase server. 
That would be awesome! There’s still much to do but I hope it improves fast. Of course an own provisioning on internal hosts is entirely possible as well. 


PS: Also remember to set stuff like limit for files, net.ipv4.ip_local_port_range, multiple IPs if any etc. Otherwise you won’t go abovee 64k sockets etc. 
Yep, this is not yet done but I know that I probably need to do it soon. Some advise there wouldn’t be bad though ;)





-Cheers! 
/Steffen 


On 02 Oct 2014, at 18:19, Michael Weibel <michael.weibel+xmpp at gmail.com> wrote: 

> Hi XMPP enthusiasts 
> 
> TL;DR: Check out https://github.com/mweibel/xmpp-server-benchmarks and contribute. 
> 
> Choosing a server for XMPP is hard, even more so as we have more and more implementations available. 
> 
> I’d like to help in choosing the server based on statistics of the scenario you imagine for your use case. 
> Using tsung [1] and ansible [2] I started with an automated XMPP server benchmark tool [3]. After making it work on Vagrant (local testing), TravisCI (automated validation of contributions) and EC2 (running the real test) I’d like to announce it publicly and invite contributors to help me (and the community as a whole). 
> 
> My goal is to have a suite of multiple servers, probably in multiple configurations, and multiple scenarios to test specific use cases (max users connected, max websockets perf, max users in MUC, max messages exchanging etc.). 
> 
> Please check it out, contribute servers (only mongooseIM and ejabberd as of now) and especially scenarios. Help me making the suite more efficient and making the statistics readable in some way. 
> I plan to run those tests regularly on EC2 as long as I can support it with my money (or as long as I still have a free AWS account.. ;)). 
> 
> Of course, a benchmark is always flawed in some way and should not be the only reason for choosing a server. It might however help for giving another reason for or against a certain server in a certain scenario. Also it might help server developers in discovering where they could improve. 
> 
> I’m eager to read your comments and happy to answer any questions. 
> 
> Best, 
> Michael 
> 
> 
> [1]: https://github.com/processone/tsung 
> [2]: http://ansibleworks.com/ 
> [3]: https://github.com/mweibel/xmpp-server-benchmarks 
> _______________________________________________ 
> JDev mailing list 
> Info: http://mail.jabber.org/mailman/listinfo/jdev 
> Unsubscribe: JDev-unsubscribe at jabber.org 
> _______________________________________________ 

_______________________________________________ 
JDev mailing list 
Info: http://mail.jabber.org/mailman/listinfo/jdev 
Unsubscribe: JDev-unsubscribe at jabber.org 
_______________________________________________ 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://www.jabber.org/jdev/attachments/20141002/1e973816/attachment-0001.html>