[Gllug] Replacement SATA disk naming

John Hearns hearnsj at googlemail.com
Thu Mar 5 12:49:23 UTC 2009


2009/3/5 Alain Williams <addw at phcomp.co.uk>:
> >
> [**] System new in November, a disk (Seagate Barracuda) failed/replaced in Jan
> and this new disk (MAXTOR) failed a couple of days ago.
> A different m/c with new 1TB disks had one of them fail after 3 weeks.
> Is 1TB pushing the technology limits or am I just unlucky ?

I would say in some senses yes.

Let me explain - it may make sense for desktop machines to have 1Tbyte
or even larger disks - so the PC Worlds and Currys of this world can
say "Hey - Look! Buy a HUUUUGEE disk for all your
photographs/music...yadda yadda"

However, in the SERVER market let's take a step back. Why are we going
for HUUUGEE single disks? It is inviting failure.
I would say no - a small(ish) disk for the OS, which is hopefully more
reliable, and use those big disks for data storage.
If your OS disk goes phut then it is a bigger pain than restoring data.
And remember if you have a RAID5 the reconstruction time goes up with
the disk size.


In the Beowulf/HPC field there is a lot of talk of servers being
equipped with a flash drive for the OS. I think this will happen more
and more - a sensibly sized flash device for your root filesystem, and
the applications /usr /usr/local mounted from a SAN or a NAS server.
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug




More information about the GLLUG mailing list