[Devel] follow up to the comparing test for SER and OpenSER from TransNexus

Daniel-Constantin Mierla daniel at voice-system.ro
Tue Jun 19 17:11:27 CEST 2007


Hello,

On 06/12/07 04:15, Di-Shi Sun wrote:
> All,
>  
> This e-mail is follow-up to our previous posting about tests we 
> performed at TransNexus comparing SER and OpenSER.  Our tests 
> were for wholesale VoIP carriers who want a straightforward, unbiased 
> comparison of OpenSER and SER.  As part of this goal, we purposely 
> chose the simple implementation of building each software with default 
> settings.
>  
> Folks from the SER community have pointed out to us that building each 
> platform based on the default configuration does not lead to 
> an accurate comparison because debug and blacklisting settings for SER 
> and OpenSER are different and favor the performance results of OpenSER.
>  
> Since our goal is to provide users with the most accurate comparison 
> possible, we are seeking feedback from both communities on what 
> changes should be made to the SER and OpenSER compile/configuration 
> options in order to measure an accurate comparison between the two 
> software platforms.
>  
> We are considering re-running the performance test with debug and 
> blacklisting turned off for both OpenSER and SER based on advice 
> from folks in the community that this configuration will produce a 
> more accurate comparison.
>  
> Does anyone else have any advice or input to offer before we re-run 
> the Performance Comparison test?
it is quite hard to define similar parameters in both software. 
Attention has to be paid to default values for sensitive parameters like 
time outs, hash table sizes, debug levels, memory and locking, number of 
processes, a.s.o.

This will need to investigate quite a lot of source code. Perhaps, if 
the difference is dramatically, then worth to investigate, if there is 
just a bit, it is clear that depending on who is doing the config can 
get better results for some cases. The way config file is written in 
each case can have big impact on final results.

Cheers,
Daniel

>  
> Thank you,
>  
> Di-Shi Sun.
>  
> ------------------------------------------------------------------------
>
> _______________________________________________
> Devel mailing list
> Devel at openser.org
> http://openser.org/cgi-bin/mailman/listinfo/devel
>   



More information about the Devel mailing list