Dexter Filmore wrote:
The SCSI command set allows more precise fault detection. What good is a raid
if the system deosn't know about failed disks?
Umm, really? Reference?
I thought that S.M.A.R.T. worked for both ATA and SCSI drives and was as
good as it gets.
Or are you talking about SAF-TE? That's part of the IPMI standard and
is just for enclosures.
I'd rather go for harware less likely to fail n the first place. The one thing
is "precaution", the other is "lessen risk and damage in fault case".
I like precaution better.
That's fine. You are willing to pay the cost premium to get that.
Some would rather take the cost premium and allocate it to a different
characteristic like total storage size. That's fine too.
Same reason why don't prefer a shoddy car with airbag over a decent car with
airbag solely for a bargain.
That's a poor analogy. What you are protecting in that can't be replicated.
Yet ... ;)
Comes down to "depends" - how much your data and uptime is worth to you and
how critical it is in a produvtion environment.
Uh, no. Now that's just blather.
Both ATA and SCSI storage *systems* can be made equally reliable. In
fact, ATA systems can generally be made more reliable for less money
than SCSI in most cases. For example, I can generally run RAID-10 ATA
with a hot spare for the same price as RAID-5 SCSI for the same amount
of storage. That means I can take ln(n) disk failures (n/2 disk
failures best case, but probability suggests that if you let it go that
long you will get a duplicate hit) rather than just 1 disk.
You may not like the fact that individual parts fail more often because
you don't want to mess with swapping out the drives. That I can
understand since very few people are willing to pay for a proper
hot-swap enclosure and put in place the procedures to cope with it.
However, properly run companies assume that parts will fail *regardless*
and have procedures in place. Normally, they set up a contract with
someone like Network Appliance.
-a
--
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
