As one of the others to ask this question, here are my findings so far:

- Most people aren't keeping their LTO3 happy, some even falling victim 
to the marketing crap that "LTO technology isn't harmed by shoeshining." 
I have personally seen a 20-25% (per month) drive failure rate in an 
environment of constant shoeshining on LTO1.
- Software striping of multiple LUNs may be a good quick fix or stop gap 
solution.  Especially if you can aggregate your FC interfaces this way.
- A few have mentioned that they started using FC disk because SATA was 
too slow.
- When buying, pay careful attention to how your storage array does 
parity and whether it can read down all paths in a RAID-5 setup.  I have 
personally had much better success on the STK (Engenio) FLX210 (SATA) 
arrarys for read performance than I did from EMC FC4700 (FC), though not 
an apples-to-apples comparison, the bottleneck appeared to be in the 
controller, and not the front- or back-end interfaces.
- One of the sites I manage uses DSSU with LTO1 because all of their 
backup clients are way too slow and even LTO1 is overkill.  For them, I 
just throw more disk than I need at it and make sure the destage window 
doesn't overlap the backup window.  You may need to get creative, for 
example: de-stage 6a-6p every 10 hours, that way the noon-3p window 
doesn't interfere (since NBU won't let you do two windows in the same 
day).  Even with a 24x7 window, you could achieve this with multiple 
policies although it may get hairy and is admittedly a hack.

I think you're best bet is to do some software striping today, and demo 
a few vendors' offerings for your next disk platform.  Don't rule out 
FC.  I'm not saying you'll need to go with a high-end array, but 
possibly not SATA.


----Original message----
Date: Fri, 22 Sep 2006 10:19:15 +0100
From: "Weber, Philip" <Philip.Weber at egg.com>
Subject: [Veritas-bu] LTO3 and Disk Staging

Sorry if this has been done to death recently but...

What sorts of disk systems are people using as disk staging to keep
LTO3s happy?  Our current EMC Clariion Cx300 struggles to supply data
fast enough to 3 LTO2 drives (even outside of NetBackup) and performance
goes through the floor if (as is often the case) it is still trying to
write backups to disk while streaming data off to tape.  We have pretty
much a 24*7 backup window with lots of slow & small, slow & big, and
some fast, clients.

It seems the idea that you need disk staging to keep your LTO tapes
running nicely, rings a bit hollow.  Also seems to me we're looking at
needing a high-end disk system.  Which makes a mockery of backing up to
cheap ATA/SATA disk being the way forward.

The only way I can see to get around the performance problems with
writing/reading backup streams to/from disk concurrently, other than a
high-end system, is to script some kind of mirror-split-off process and
use it to do the duplication with scripts rather than DSSU.  i.e. back
up to disk overnight & then split a mirror off (to somewhere where its
I/O will be separate) and duplicate off this during the day before
resynching.  Anybody do anything like that?  I can see all sorts of
nasty gotchas ... maybe that way madness lies?

cheers, Phil



-- 
Joe Royer / SysAdmin / Digital Motorworks / 512-692-1028


This message and any attachments are intended only for the use of the addressee 
and may contain information that is privileged and confidential. If the reader 
of the message is not the intended recipient or an authorized representative of 
the intended recipient, you are hereby notified that any dissemination of this 
communication is strictly prohibited. If you have received this communication 
in error, please notify us immediately by e-mail and delete the message and any 
attachments from your system.