[Beowulf] Consumer vs. Enterprise Hard Drives in Clusters

Eugen Leitl eugen at leitl.org
Thu Jan 22 02:54:05 EST 2009


On Wed, Jan 21, 2009 at 04:19:55PM -0800, Jon Forrest wrote:

> One issue I always face when buying RAID storage is whether
> to get the more expensive "enterprise" drives or to go
> ahead and settle for the consumer grade drives. The
> last time I checked, the enterprise drives were ~25%
> more expensive than the consumer drives. This difference
> might have changed since then.

It depends on whether you go for same-size SATA,
or 10 krpm SATA, or 15 krpm SAS. Very large price
differences there; of course it's comparing apples
and oranges.
 
> I've tried to find out what the differences are between
> the enterprise vs. the consumer drives. I've heard
> different answers. One answer I've heard is that the
> enterprise drives use "better" firmware than the

Enterprise drives are typically better equipped to deal
with vibrations in large RAIDs (and an enterprise drive
can perform worse than equivalent consumer SATA drive
at same workloads). This is exacerbated at
high densities. Their raw error rate is lower. The SAS
versions have other goodies.

The drives are actually reasonably sensitive:
http://de.youtube.com/watch?v=tDacjrSCeq4

> consumer drives. I wasn't sure whether I should accept
> this explanation, especially since those famous papers
> from Google and CMU said that they didn't see any
> significant differences between the two types of drives.

One interesting thing is that they said drives died sooner
at temperatures usually recommended for data centers, since
they were age-tested at higher temperatures, so bias crept
in. I've seen that anecdotally myself.

I would like to see the same data for Raptors, Velociraptors
and SAS.

> Also, since (until recently) the manufacturer's warranty was
> the same on both drive types, I was skeptical about the
> differences between the drives.
> 
> In the last week or so, it has come to light that Seagate
> has a big problem with their 1TB disk drives (and others).
> (See 
> http://www.computerworld.com/action/article.do?command=viewArticleBasic&taxonomyName=storage&articleId=9126360&taxonomyId=19&intsrc=kc_top
> and 
> http://forums.seagate.com/stx/board/message?board.id=ata_drives&thread.id=3668
> for coverage of the problem)
> What's interesting about this is that *both* the consumer
> (e.g Barracuda 7200.11) and enterprise (e.g. Barracuda ES.2)
> versions of the Seagate drives have the same firmware problem,
> or at least are suffering from the same firmware bug.
> This makes me wonder about how different their firmware must be.

IIRC the official explanation is that there's a zero counter issue
triggered by some measurement devices, which renders the drive
inaccessible when switched on. It can still be recovered with 
appropriate (difficult to find out, since, as frequently, there's
the same serial for a wide range of devices, and only Seagate
has databases which know for sure).

I'm not sure I buy this explanation, since I've seen a lot of
creeping SMART issues with that product line, and I've had a
lot of drives fail while on. Maybe they're having different issues.

While every producer has had a bad batch I'll think I'll stick
with WD drives for a while. I'm really looking forward to SSDs,
especially when they move on to a MRAM or similiar technology.
 
> I'm glad I've been buying the consumer drives because, at least
> as far as this problem is concerned, my life wouldn't have been
> improved if I had spent the extra money for the enterprise drives.

If you're just using these for scratch space, and semi-sequential workloads
in vibration-poor environments, yes.

-- 
Eugen* Leitl <a href="http://leitl.org">leitl</a> http://leitl.org
______________________________________________________________
ICBM: 48.07100, 11.36820 http://www.ativel.com http://postbiota.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE
_______________________________________________
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.



More information about the Beowulf mailing list