[TriLUG] Re: New system configuration thoughts
Kevin Flanagan
flanagannc at gmail.com
Thu May 3 19:17:12 EDT 2007
Reliability/availability:
In installations where we would have a mirror for the OS, and one or more
RAID 5 sets for data I would assign one additional hot spare to all arrays.
Hardware RAID will rebuild a good bit faster than a software RAID
configuration in most cases.
Performance:
RAID 4 is really among the fastest, but no redundancy, RAID "10", or "50"
are fast, but very expensive in overhead.
Hardware:
SATA has not been as reliable, perhaps due to the focus on the
desktop/consumer space. I think that is becoming less of an issue as time
goes on. Serial SCSI is new technology, very fast, and is implemented in
ways that seem to be a bit odd to those of us who are used to the "old
ways". As I understand it in servers there are more limitations on how they
can physically fit the disks in the box, rather than the old Parallel SCSI
supporting 7 devices per buss, SSCSI supports so many more disks that HP has
chosen to break up the number of initiators in the box in seemingly odd
ways.
So many option, so many details, so much cost......
Kevin
On 5/3/07, Sean Myers <smyers at americanri.com> wrote:
>
> > Higher numbers of drives means smaller drives means it takes less time
> > to rebuild a failed drive onto a hot spare, thus shortening the
> > vulnerability window.
>
> I'm not seeing how the number of drives you have in an array determines
> the size of those drives. Higher numbers of drives does not mean smaller
> drives in an array.
>
> --
> Sean Myers
> System Administrator
> American Research Institute
> (919) 228-4961
> --
> TriLUG mailing list : http://www.trilug.org/mailman/listinfo/trilug
> TriLUG Organizational FAQ : http://trilug.org/faq/
> TriLUG Member Services FAQ : http://members.trilug.org/services_faq/
>
--
+---------------------------------------------------+
Fear and hatred clouds our judgment
Free us all from endless night
John Lennon from "Attica State"
More information about the TriLUG
mailing list