On Tue, 6 Oct 2009, John Stoffel wrote:
> Fun!

One of the better jobs I've had.  :)

> I like Netapp too, but I'm getting more and more pissed at their silly
> 16Tb limit on aggregates, it really starts killing you when you have
> 12 shelves, each with 14 1Tb drives in them.  It makes carving up
> storage painful again.

I agree -- it's been a longtime beef of mine that they insist on limiting 
capacities based on what they think I want.  I wish they would allow me to 
exceed their limits with the understanding that it could affect 
performance -- especially when I know I don't need performance in favor of 
large archives and such.

> a) backup the data from the backend, completely away from the frontend
>   which is fast for backups, but makes restore horribly painful.
>
> b) backup *through* the frontend, which blows your caching unless
>   there's a way to say that all access from host X,Y or Z doesn't affect
>   the cache.  It's also slower, since you're now putting the load all on
>   the frontend, which is where all your clients work through as well.

Interesting points.

> So I'm pretty down on NFS aggregation boxes like Acopia, unless they
> can address the backup issue well, and I haven't seen that yet, but I
> also admit we've given up on them.

I think NetApp's GX (OnTAP 10, when it happens) will be very good for them 
if they can get their act together and make it happen.

We had Isilon in today to talk about their stuff, and BlueArc will be 
coming in next week.  Exanet hasn't returned my call yet, but I'm 
particularly curious about them.

We've recently done some Oracle tests and proved that a mediocre NetApp 
can hold its own against some pretty beefy SAN hardware, which was cool. 
I expected our humble little 3020 to fall short, but in some tests it 
outperformed the SAN.  It blew away the SATA SAN disks, and did reasonably 
well against the FC SAN (it's an FC filer).  Of course, there hasn't been 
any serious performance tuning on either the NAS or SAN.

The Isilon design looks interesting, and we're familiar with the concept 
because of XIV, which we had in house for a short time.  Isilon is not 
really ready for the enterprise, though.  They looked at us funny when we 
started talking about locking shares for compliance, and they don't do 
cloned or writable snapshots.

I was surprised to hear that they do copy-on-write snapshots.  I'm curious 
to see one in action to see how the performance is affected by that.  My 
guess is that the reason they can do it is because of their massively 
parallel hardware.  If they were stuck with just a couple of controllers, 
the performance hit of copy-on-write would kill 'em.

-Adam

_______________________________________________
Discuss mailing list
[email protected]
http://lopsa.org/cgi-bin/mailman/listinfo/discuss
This list provided by the League of Professional System Administrators
 http://lopsa.org/

Reply via email to