network performance

[Stefan Lambrev]

Greetings,

Kris Kennaway wrote:
> Stefan Lambrev wrote:
>> Greetings,
>>
>> Kris Kennaway wrote:
>>> Stefan Lambrev wrote:
>>>
>>>>> I'll use again hwpmc and LOCK_PROFILING to see what's going on.
>>>>> And will try the same benchmark on quad core processor as now 
>>>>> numbers of cores/cpus matter :)
>>>>>
>>>> Here are promised results - http://89.186.204.158/lock_profiling-8.txt
>> Finally I got pmcstat working - http://89.186.204.158/hwpmc-p4.txt
>> The stats are gathered during 600kpp incoming.
>> I think that syncache or what calls MD5Transform is not SMP able, and 
>> that's why outgoing 250kpps is the limit that I can't beat.
>
> It looks like the syncache is using most of the CPU time.  However you 
> are not hitting problems caused by lack of concurrency there.  It does 
> do a *lot* of work with the syncache mutex held (including generation 
> of the cookie, which involves MD5) so it might be an issue in the 
> future, but there are other bottlenecks in the way before that is your 
> main issue.  Things may be different with more CPUs.
>
> Did you compare to what happens to performance when the syncache is 
> disabled?
>
> Kris
>
When I disable syncookies the server respond to more packets - from 
250kpps enabled to 320kpps when disabled.
Can I disable syncache and how?
I'll try to increase syncache limits and will test again.

Btw is this expected - net.inet.tcp.syncache.count: -387.
I think this number should be always > 0.


-- 

Best Wishes,
Stefan Lambrev
ICQ# 24134177