On Thu, 25 Nov 1999, Michail D. Flouris wrote:
> Hi there,
>
> I'm using polygraph 2.2.4 on a Solaris 2.6 machine to test
> some research ideas and I was wondering about its traffic
> generating behaviour.
>
> When I run polygraph and squid on the same machine and
> I leave the robots get the "best effort" approach
> (I use no "req_rate") the performance I get over time
> is not around some constant rate ...
> For example, for the first minute I get around 50 reqs/sec
> then for the second minute I get around 2 reqs/sec and the machine
> is idle, then I get again my 50 reqs/sec then again 2 reqs/sec,
> and so on ...
>
> Can someone please explain why this happens, and how can I make
> the robots constantly make requests and not lying idle during
> any period of the benchmark ?
>
If you run netstat -na - you will see lot of connections in TIME_WAIT
state.
Decerese TIME_WAIT interval using
ndd -set /dev/tcp tcp_close_wait_interval 60000
or even 30000
Igor Khasilev |
PACO Links, igor@paco.net |
This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:09 MDT