Re: [squid-users] Veritas filesystems and Squid

On Fri, 17 Aug 2001, Deb Heller-Evans wrote:

> Does any have of any known performance issues using Veritas disk
> management (and thus, filesystems) with SQUID on Solaris 2.7?
>
> Also, in reading the documentation on the web, I saw somewhere
> recommended that RAID5 *not* be used as the target disk(s) for
> the cache. In trying to maintain uptime, should a disk go bad,
> using RAID5 would give us a WIN situation. What are the reasons
> to NOT use RAID5, and do you think they outweigh keeping a
> system up (ie, RAID5 allows for the hotswapping of disk drives
> should it become necessary)?

because you're doing essentially multiple threads of random access (both
reads and writes) dividing you i/o requests across multiple discreet
disks or stripes will generally result in higher throughput (more random
objects served/cached) than having a single cache dir/disk/stripe which
all your i/o must be served from...

in addition, there are a couple of other considerations...

your data, (cache objects) is generally not all the valuable... you can in
most cases, just unmount the disk, comment it out of your squid.conf and
keep going with your remaining disks... If you have the ability to hotswap
the disk for another, which you can then format remount, and add back in
that makes things easier, but if you have four cache disks for example you
can probably live with three for some time if you have to...

raid 5 in software, requires additional cpu overhead in order to generate
the parity information that reflects the data... with a hardware
controller this may not be an issue...

in a raid 5 system, recreating the information onto the new disk(or
preferably a hot spare if there is one) from the remaining blocks and the
parity info, can result in a serious performance hit when it happens or
on some systems may require that you take the stripe offline. inserting a
new disk, bringing it online and formatting it is probably faster...

> Thanks,
>
> Deb
>
>
>
> "If it dies, it's biology. If it blows up, it's chemistry,
> and if it doesn't work, it's physics."
> -- University bathroom graffito
> ô¿ô
> ~
>

-- 
--------------------------------------------------------------------------
Joel Jaeggli				       joelja@darkwing.uoregon.edu
Academic User Services			     consult@gladstone.uoregon.edu
     PGP Key Fingerprint: 1DE9 8FCA 51FB 4195 B42A 9C32 A30D 121E
--------------------------------------------------------------------------
It is clear that the arm of criticism cannot replace the criticism of
arms.  Karl Marx -- Introduction to the critique of Hegel's Philosophy of
the right, 1843.