RE: [squid-users] squid with linux or freeBSD?

From: Lightfoot.Michael <Lightfoot.Michael@dont-contact.us>
Date: Thu, 1 May 2003 13:41:47 +1000

> > why raid your disks at all? you'll get before performance
> out of having
> > seperate cache dirs on each disk than you will out of any kind of
> > stripe...
>
> because writing or reading a single data across multiple
> disks/stripes is
> much better than having a separate cache dirs... raid 0 or stripping
> distribute the load and minimize the positioning of a disk
> head (seek time)
> which is the most obvious bottleneck of a disk...
>
In the case of squid, it actually doesn't. Most cacheable squid objects
are small (< 2KBytes). They often only occupy a single disk block (<=
512Bytes). With multiple cache_dirs you are allowing squid to manage
its own "striping" which it does quite well thank you.

Three years ago I ran some medium-sized (30GByte) caches which I
configured as raid 0 stripes across 4 disks for the first two and then
decided, after comments on this list, that I would try 4 separate
cache_dirs for the second two (which shared the load of the first two
respectively). I found there was no discernible difference in the
performance of the two pairs and subsequently changed the original pair
to separate cache_dirs.

I suspect that the overhead of the raid software layer (in that case it
was Veritas Volume Manager) cancelled any benefit of the striping for
these small reads and writes. The size of the Solaris 8 file cache was
much more significant (ie if you throw memory at a squid proxy system
you get significant improvements in I/O reads, whether you let the OS
handle it or give a goodly amount to cache_mem.)

Michael Lightfoot
Unix Consultant
ISG Host Systems
Comcare
+61 2 62750680
Apologies for the rubbish that follows...
------------------------------------------------------------------------
NOTICE: This e-mail message and attachments may contain confidential
information. If you are not the intended recipient you should not use or
disclose any information in the message or attachments. If received in
error, please notify the sender by return email immediately. Comcare
does not waive any confidentiality or privilege.
Received on Wed Apr 30 2003 - 21:42:26 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:15:46 MST