On Tue, 15 Jun 1999, gyan prakash wrote:
> It is able to handle around 100 - 120req/sec just for
> 10 minutes not more than that.
What response times are you seeing? If your response time is X sec and
you are doing Y req/sec, then your OS will have to handle at least:
(X+60)*Y
concurrent connections in various states. Squid will have to deal with
(X*Y) connections. You probably should add about 10-20% for workload
peaks.
From my experience, things get exponentially worse on a FreeBSD with
number of connections beyond ~2-3K. Do not know if that is true for
Linux.
Alex.
Received on Tue Jun 15 1999 - 10:57:36 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:46:53 MST