Abnormal page faults and TCP connection ratio

From: Azfar Kazmi <[email protected]>
Date: Tue, 03 Nov 1998 09:44:12 +0500

I am using Squid 1.NOVM.18 on a Red Hat 4.1 Linux [kernel 2.0.34.] The
machine is a Pentium 200MHz based Digital PC 3000 with 32MB RAM. Total disk
is 2.1GB. Free disk space at the moment is around 1100MB. Following
configuration squid is currently running on:

---------------------------------------------------------------
Parameter Value Description
---------------------------------------------------------------
VM-Max 8 Maximum hot-vm cache (MB)
VM-High 90 High water mark hot-vm cache (%)
VM-Low 75 Low water mark hot-vm cache (%)
Swap-Max 300 Maximum disk cache (MB)
Swap-High 95 High water mark disk cache (%)
Swap-Low 90 Low water mark disk cache (%)
Neg-TTL 300 TTL for negative cache (s)
ReadTimeout 900 Maximum idle connection (s)
ClientLifetime 5400 Lifetime for incoming HTTP requests
CleanRate 31536000 Rate for periodic object expiring
HttpAccelMode 0 Is operating as an HTTP accelerator

With these values, the ratio between page faults and TCP connections is
very high. Page Faults divided by TCP connections always remain around
1.4-1.6. Can someone suggest me whats wrong. Due to this abnormality, my
squid machine usually extremely slows down. So much that it takes minutes
to run Pine on it!! I first had specified 16MB VM-Max but, as suggested by
the FAQ in such conditions, I decreased them to 8. Did I do right? What
parameter should I change?

Azfar
Received on Mon Nov 02 1998 - 22:47:59 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:42:56 MST