Re: Ata over Ethernet

Good points, and yes, I had thought of them. If you want reliability, you
still want redudency. The array we set up will have to be (at least) RAID 5.
I wish linux software RAID supported RAID 6 ("two dimensional" RAID, requires
N+2 drives). That way, two drives could fail, and we'd still be OK. I've
also thought of doing three RAID 5 arrays, and then further putting them into
a RAID 5 array (A RAID of RAIDs), but that seems quite the overkill.

But thanks for pointing that out...it's a good reminder.

j----- k-----

On Wednesday 27 October 2004 19:02, Jim wrote:
> When one is working with large numbers of devices, one always hits the
> MTBF wall. Even if an individual device has 10,000 hours, if you have
> 10,000 of them, you will be replacing one every hour. After talking to
> the folks running a very large compute cluster with thousands of
> processors and terabytes of memory, they replace hardware daily. The
> trick is to be able to stay in production while this is done. In the
> case of the cluster, the run does not depend on the survival of any one
> node. In the case of storage, the key is raid, hot swap, and online
> spares. If you shove in a few 250GB sata drives a year, so what? Price
> of doing business! The important thing is staying on line.
>
> My $.02
>
> Jim
> ---------
> To unsubscribe, send email to <aklug-request@aklug.org>
> with 'unsubscribe' in the message body.

-- 
Joshua Kugler
CDE System Administrator
http://distance.uaf.edu/
---------
To unsubscribe, send email to <aklug-request@aklug.org>
with 'unsubscribe' in the message body.