Re: [squid-users] CPU utilization performance issue

From: Henrik Nordstrom <hno@dont-contact.us>
Date: Sat, 9 Aug 2003 09:30:26 +0200

On Saturday 09 August 2003 01.20, Adam Aube wrote:

> How can RAID0 have worse performance than RAID5? RAID0 was
> designed to optimize disk write performance by striping writes
> across multiple disks.

Because it does not add any redundancy or performance to a Squid
setup. RAID0 only adds drawbacks compared to separate drives.

> I would think that RAID0 would at least outperform RAID1.

One cache_dir per drive gives maximal performance and flexibility for
Squid but no automatic fault management. If you complement this with
some software which removes the cache_dir from squid.conf and
restarts Squid if a drive should fail then you do get quite
acceptable level of fault recovery however.

RAID0 is on performance level close to using one cache_dir per drive,
but has the drawback that if one drive fails the whole RAID0 set of
drives is gone.

RAID1 adds redundancy, but nearly doubles the number of drives needed
for the same performance compared to separate drives.

RAID5 is slower than RAID1 for Squid, but if you don't really need the
speed but you mus have redundancy then it may be acceptable. However,
given the low price of harddrives there is barely no motivation to
use RAID5 instead of RAID1 for Squid.

My recommended base setup is a RAID1 setup for OS + logs, then
separate drives for as many cache directories you need. For low-end
setups the RAID1 set may also be used for cache.

On high-end setups using RAID1 for the cache drives is recommended
method of increasing the reliability and decreasing management cost
at only the cost of doubles number of drives needed for cache.

-- 
Donations welcome if you consider my Free Squid support helpful.
https://www.paypal.com/xclick/business=hno%40squid-cache.org
If you need commercial Squid support or cost effective Squid or
firewall appliances please refer to MARA Systems AB, Sweden
http://www.marasystems.com/, info@marasystems.com
Received on Sat Aug 09 2003 - 01:31:23 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:18:48 MST