On Tuesday 25 September 2001 10:19 am, Peter Smith wrote:
> I'm currently running Squid-2.4.stable2 on a dual 866mz P3, 512mb RAM,
> scsi raid cache dir, Redhat 7.1 (2.4.2-2smp) . I'm seeing squid run on
> only a single processor while I have about 200+ users hitting it. Is
> there anything that I might be doing wrong? I'm using the following
> during configuration of the squid source:
A few thoughts:
1. Definitely update your kernel. If you can't bring yourself to use a
non-RedHat kernel, at least installed the v2.4.3-12 that RH released some
time ago. (See ftp.redhat.com/pub/redhat/linux/updates/7.1/en/os)
2. How do you know that only a single CPU is being used? I mean, what
tool is being used to report the CPU use?
3. From your </configure --enable-storeio="ufs,null"> it seems that you are
not building the multi-threaded store types (aufs, diskd) into Squid. Why
not?
4. RAID can boost your STR significantly, but it doesn't do anything for
disk access times. Since the files cached by Squid tend to be smallish, it
is the access time (rather than STR) that really helps Squid performance.
Received on Tue Sep 25 2001 - 10:05:25 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:02:29 MST