RE: [squid-users] Raid 0 vs Multiple stores.

> -----Original Message-----
> From: Rabie van der Merwe [mailto:Rvdmerwe@mhg.co.za]
> Sent: Thursday, 28 November 2002 9:11 PM
> To: squid-users@squid-cache.org
> Subject: [squid-users] Raid 0 vs Multiple stores.
>
>
> If I have a box with the OS on a disk, and I
> have to choose between running 4 disk on a
> separate SCSI channel and running a store on
> each disk vs a RAID controller with a battery
> backed up cache running also 4 disks striped
> with write-back caching enabled, in my mind
> the raid controller should be the faster option?
>
> Has anyone benchmarked such a scenario?
>
I haven't benchmarked this particularly scientifically but a few years
back at a previous site we first configured a 32 Gbyte cache on a 4 x 9
Gbyte disk simple stripe (Solaris 2.6, Veritas Volume Manager software.)
After this cache had filled and run for a while we configured another
cache at the same level in our hierarchy (as an ICP sibling peer for
redundancy and load sharing) but as four separate 8 Gbyte cache_dirs, as
simple disks with no VxVM control.

We found that the second actually performed slightly better as well as
providing cache_dir level redundancy. I assume the performance
difference was due to the overhead in running VxVM, or it may simply
have been statistical variance. We changed all our multiple disk caches
to use simple disks after that (mostly to utilise cache_dir redundancy.)

Michael Lightfoot
Unix Consultant
ISG Host Systems
Comcare
+61 2 62750680

------------------------------------------------------------------------
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 Thu Nov 28 2002 - 18:08:38 MST