Veritas-bu

[Veritas-bu] multiplexed duplicates & vaulting

2001-10-31 12:48:37
Subject: [Veritas-bu] multiplexed duplicates & vaulting
From: Jason.Ahrens AT telus DOT com (Jason Ahrens)
Date: Wed, 31 Oct 2001 12:48:37 -0500
> -----Original Message-----
> From: David A. Chapa [mailto:david AT xbpadm-commands DOT com]
> Sent: October 31, 2001 11:03
> To: Steve Mickeler
> Cc: Veritas BU
> Subject: RE: [Veritas-bu] multiplexed duplicates & vaulting
> 
> 
> Well let me be so bold as to ask the question...Is MPX 
> really necessary on the front-end?
> 
> I usually MPX only incremental backups, FULLS I 
> recommend that they do not MPX.
> 
> Just a thought.

Good thought, but there are a few problems if I don't mpx my backups.

As we are in a full 24x7 environment here, a restore request can come
through literally at any time, and has to be serviced just as fast. By
mpxing my full backups I achive a faster time-to-restore-start if (worst
case) full backups have just been queued and started. I get to a point where
the restore can be serviced sooner. As well, certain NT systems take a very
long time to backup, and at that they trickle out data (very busy systems).
By mpxing I don't loose the drives extra capacity (ie: drives can write at
10MB/s, some systems only feed data at 2-4MB/s, leaving another 6-8MB/s (60
to 80%) unused and wasted capacity). As our environment grows we find we
have to fit more and more clients into the same backup window, and mpx helps
greatly with that. We seem to be getting improvements of 20-30% (overall)
time savings using mpx on fulls. The time/window issue is the other reason
I'm not keen on dedicating a tape drive only for restores. Loosing the drive
for backups has obvious implications for the length of time to perform them.

I guess it's all in the tradoffs.

Jason

<Prev in Thread] Current Thread [Next in Thread>