Hello Henrik,
I am reading Squid code. My squid 24s1 is becoming bottleneck in cpu usage, found that most of the cpu was with poll calls. I have read this from one book that we can set SO_SNDLOWAT socket option to high value so that we can reduce the number of read calls. Does this make sense?
And another thing, I have seen in cachemgr that no of counters for select is such and such. Is there any calculation to check whether Squid is in Normal state like No of requests Vs No of poll calls, etc, etc ????
Again, What are the issues when we make Squid as a forking a child process per request? Will it not be suitable for Squid?
Thanks for any help,
Ronald
Received on Fri Aug 17 2001 - 10:58:32 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:01:42 MST