Notes on the Accusys ACS 75170 RAID box
Introduction
The ACS75170 is the SATA version of the already reviewed
ACS7500.
As they are very similar in many respects, only the differences
between the ACS7500 and the ACS75170 will be discussed here.
Design and Operation Notes
- The box looks identical to the ACS7500's, however the trays
are obviously different, internally.
- Keys are compatable between the ACS7500, ACS7630, ACS76130 and
ACS75170.
- The rebuild rate is SUBSTANTIALLY faster than the 7500 systems,
approximately 50G/hr (vs. 10G/hr). This is great.
There may be a greater impact on disk performance.
- Unlike the ACS7500, but like the ACS7630, the ACS75170 "chirps"
during the rebuild every minute or so. This sucks.
There is no reason for this -- rebuilding is a healthy process, and this
makes the thing annoying to co-workers.
- Power connector on the back of the box is NOT the normal SATA power
connector, but rather the traditional hard disk connector.
In fact, there are TWO power connectors on the back, wired in parallel.
Either can be used, Accusys recommends using both, which is not a bad
idea if you have the extra power connectors laying around.
I suspect a lot of people will love this decision, and a lot will hate
it.
Depends which you have more of laying around.
Real world recovery
My employeer recently deployed about 20 of these boxes to our branch
offices recently. Our results so far: three disk failures, and
very good results from the Accusys boxes.
The disks we are deploying were some brand-new 300G drives, and not
unexpectedly, they have started failing.
One drive caused us some weird problems, in that it apparently would
start massive retries on reading data until it finally did succeed
in a good read, so the Accusys box didn't "fail" the drive, and the
system performance was absolutely horrible.
However, the Accusys box made it very clear which drive was being
waited for, so unlocking that drive caused the system to "perk up"
to full speed, and ran fine until we replaced the disk.
The other two, however, were simple "Drive goes bad, Accusys box
starts beeping, users keep working, replace drive and system remirrors"
just the way it should be.
One thing we did discover is that our primary app for these things is
so horribly disk intensive (massive thrashing across a 70+G database)
that the performance during remirror was basically unusable.
So, what we do after a disk failure is ship them a new disk in an
Accusys carrier Next Day Air, and they insert it just before going
home for the evening.
User response has been fairly positive -- we have had no problem walking
people through the replacement process over the phone (granted, we do
ship out the drive IN the carrier), but the response at the branches
has always been, "That's it??". Yep.
Holland Consulting home
page
Contact Holland Consulting
since November 10, 2005
(C)opyright 2005, Nick Holland, Holland Consulting
$Id: acs75170.html,v 1.5 2006/07/04 20:20:35 nick Exp $