Re: [squid-users] Performance question

From: Matus UHLAR - fantomas <uhlar@dont-contact.us>
Date: Thu, 30 Jun 2005 10:00:53 +0200

On 29.06 16:29, John Halfpenny wrote:
> that's very interesting, and totally contrary to what i'd been led to
> believe (i must try to find the reference i was working from which cites
> raid 0 as an effective caching solution!)

It's mentioned in http://www.squid-cache.org/Doc/FAQ/FAQ-3.html#ss3.11

> there are a couple of performance tweaks i am going to implement on my
> next installation (i like to start again once i've done experimenting to
> make sure i get it right), so i'll disable the s/w raid 0 and create some
> separate cache directories on a dummy run to see how it goes. :)

good luck.

BTW I think I've found another reason why RAID should not behave better with
squid: even SCSI disks behave better on sequential reads. So, fetching a
file sequentially from a drive (case of multiple filesystems on multiple
drives) should be faster than fetching some parts (size of a stripe) from
one disk and some parts from second disk. So, when you are using stripes and
fetch one file, multiple disks are being used, without getting any benefit
of it.

Note that there is teoretical possibility to fetch from disks in parallel
and then join the data in memory buffers (so this would really join the
performance of both drives), but:
1. I don't know if such feature is implemented anywhere - the OS sees one
drive, so the RAID should implement it
2. when fetching multiple files in parallel, which is quite often if your
squid is really loaded, disks in RAID will have seek too much which would
decrease performance.

-- 
Matus UHLAR - fantomas, uhlar@fantomas.sk ; http://www.fantomas.sk/
Warning: I wish NOT to receive e-mail advertising to this address.
Varovanie: na tuto adresu chcem NEDOSTAVAT akukolvek reklamnu postu.
I don't have lysdexia. The Dog wouldn't allow that.
Received on Thu Jun 30 2005 - 02:00:58 MDT

This archive was generated by hypermail pre-2.1.9 : Fri Jul 01 2005 - 12:00:03 MDT