On Wed, Jan 7, 2009 at 6:43 PM, JZ <j...@excelsioritsolutions.com> wrote:
> ok, Scott, that sounded sincere. I am not going to do the pic thing on you.
>
> But do I have to spell this out to you -- somethings are invented not for
> home use?

Yeah, I'm sincere, but I've ordered more or less the same type of
hardware for commercial uses in the past.  There are a number of uses
for big, slow, cheap storage systems.  Disk-based backup is an easy
one--from a price/capacity standpoint, it's really hard to beat a
rackload of 4U systems stuffed full of cheap disks.

Not every application needs redundant power, multi-pathed disks,
highly-engineered servers, and a fleet of support engineers waiting
for your call.  In my experience, very few applications actually need
that--cheap, somewhat reliable systems with good replication and
failover usually beat "enterprise-grade" hardware anytime that the
cheaper hardware is even an option.  If you have a high transaction
rate, a need for perfect coherency and consistency, and failure is
expensive, then spending 3-10x the money for slightly higher
performance and slightly lower failure rates makes perfect sense.

Then again, I'm used to having enough quantity flying around to make
the cost differences worth it.  Spending 100 hours of staff time to
save $2k up front is dumb.  The last time I built commercial storage
servers like this, it took about two extra months of my time dealing
with vendors and qualifying hardware, but we shaved $250k off of a
$350k budget when the company was strapped for cash.  That was an easy
call.

It's all about quantifying your risks and knowing what you really
need.  In my experience, any time you can make software-based
replication do what you want, and you aren't paying massive per-server
software license fees, you're probably better off with a larger number
of cheaper systems vs. a smaller number of more expensive systems.


Scott
_______________________________________________
zfs-discuss mailing list
zfs-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/zfs-discuss

Reply via email to