ADSM-L

Re: Compression

2000-02-01 02:07:21
Subject: Re: Compression
From: "Lambelet,Rene,VEVEY,FC-SIL/INF." <Rene.Lambelet AT NESTLE DOT COM>
Date: Tue, 1 Feb 2000 08:07:21 +0100
Hello,

we use compression for every client (NT, Win95, HP-UX, Backint, Exchange
agent, AIX, Solaris), except for Netware ones (Netware data is already
compressed on the client disks).

We have to do this because we only have 100 GB of disk storage pools and
this has to take all the backups of one night.

Our backup rates are ok, probably due to fast enough client hardware.

Thanks for your contributions,

                René Lambelet
                Nestec S.A. / Informatique du Centre 
                55, av. Nestlé  CH-1800 Vevey (Switzerland) 
                *+41(021) 924 3543  7 +41 (021) 924 4589  * B 133
                mail to: rene.lambelet AT nestle DOT com 

                This message is intended only for the use of the addressee
        and may contain information that is privileged and confidential.


> -----Original Message-----
> From: Kelly Lipp [SMTP:lipp AT STORSOL DOT COM]
> Sent: Monday, January 31, 2000 11:40 PM
> To:   ADSM-L AT VM.MARIST DOT EDU
> Subject:      Re: Compression
> 
> Paul,
> 
> I agree with your observations.  However, I can't rationalize them with
> the
> lack of elapsed time backup improvement.  Yes, compressing earlier rather
> than later makes sense.  But why, then, don't we see an improvement in
> overall backup times?
> 
> I really like compression to keep the overall disk storage pool
> utilization
> low.  That's a real good reason to do it, if you don't have enough disk
> pool.
> 
> How about somebody that compresses everyplace giving us their real world
> experiences?
> 
> Kelly J. Lipp
> Storage Solutions Specialists, Inc.
> PO Box 51313
> Colorado Springs CO 80919
> (719)531-5926
> Fax: (719)260-5991
> www.storsol.com
> lipp AT storsol DOT com
> 
> -----Original Message-----
> From: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU]On Behalf Of
> Paul Zarnowski
> Sent: Monday, January 31, 2000 3:22 PM
> To: ADSM-L AT VM.MARIST DOT EDU
> Subject: Re: Compression
> 
> 
> At 04:21 PM 1/31/2000 -0500, Johnson, Chris E. wrote:
> >All,
> >
> >Is there any advantages or disadvantages to using "Compression" with how
> it
> >relates to Nodes in *sm?
> 
> I have to take an opposing view to Kelly's on compression.  I would argue
> that you should enable compression at the client, unless you have a
> particularly slow client system.  Running single-backup performance tests
> will probably not show any differences between compression and
> non-compression.  However, logically it just makes sense to compress data
> as early in the process as you can, and to distribute the overhead of
> compression out to the client systems, if possible.  Kelly may be correct
> that tape drive compression is as good as *SM compression.  However, IMHO,
> doing compression at the tape drive misses performance/thruput advantages
> that you could have gotten if doing compression at the client.  Backup
> data
> originating at a client system will take the following path in a typical
> *SM environment:
> 
> 1. Client System
> 2. Network
> 3. *SM server RAM
> 4. *SM server disk storage pool
> 5. *SM server tape storage pool (migration from disk)
> 6. *SM server tape storage pool (reclamation)
> 
> Using client compression should yield performance/thruput advantages in
> steps 2-6, while using tape compression will only yield performance
> advantages in step 6.  I don't know if using tape drive compression will
> slow down tape I/O, but I doubt that it will, so that shouldn't be a
> factor.
> 
> As Kelly says, if you have enough network bandwidth, then step 2 doesn't
> matter, but step 4 could still make a difference.  In any case, why NOT
> use
> client compression?  Most systems are fast enough now that compressing
> data
> doesn't cause any performance problems on the client system.  If you think
> you might someday run out of resource in any of steps 2-4, then you might
> as well use client compression.
> 
> my 2 cents.
> ..Paul
<Prev in Thread] Current Thread [Next in Thread>