httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ralf S. Engelschall" <...@engelschall.com>
Subject Re: cvs commit: apache-2.0/mpm README.rse
Date Thu, 12 Aug 1999 15:22:36 GMT

In article <199908121421.QAA06519@deejai.mch.sni.de> you wrote:
>>    
>>      Some quick tests with the dexter MPM and both GNU Pth and FreeBSD uthread
>>   -  showed the following not too bad results with ApacheBench:
>>   +  showed the not too bad results with ApacheBench under the following config:
>>   +
>>   +  NumServers           1
>>   +  StartThreads        32
>>   +  MinSpareThreads     32
>>   +  MaxSpareThreads     32
>>   +  MaxThreadsPerChild  32
>>   +  MaxRequestsPerChild  0
>>   +
>>   +  With more processes the results are not such good. Then both user space
>>   +  threading libs get only up to 400-500 req/sec. The reason seem to be the
>>   +  unserialized acccept problem.
> 
> Does it make sense to set MinSpareThreads to the same value as MaxSpareThreads?
> I thought that would cause unneccessary housekeeping on light load
> changes. There ought to be a "hysteresis" to allow for load changes
> without having to create/destroy threads all the time.
> 
> Or am I completely wrong?

Hmmm... yes, that can be true. It should be a little bit lower in theory. I'll
look whether it makes differences in practice. Thanks for the hint, Martin.
For the real effect we've to look inside the dexter.c to find out how it's
actually implemented there...
                                       Ralf S. Engelschall
                                       rse@engelschall.com
                                       www.engelschall.com

Mime
View raw message