ADSM-L

Re: Client Compression Rate

2015-10-04 17:51:04
Subject: Re: Client Compression Rate
From: Prather, Wanda [SMTP:PrathW1 AT CENTRAL.SSD.JHUAPL DOT EDU]
To: ADSM-L AT VM.MARIST DOT EDU
I agree with Andy, there is no guaranteed way to tell.

I have worked with several different ADSM installations.  As Andy said, text
data compresses well, executable files don't.  If you have a wide variety of
client machines, say a mixture of Unix and Windows machines, a mixture of
file servers, print severs, desktops, etc., on average you will probably see
slightly better than 2 for 1 compression, but not 3 for 1.

If your client machines are all the same configuration, say all Oracle data
bases, or all mail servers, then the best bet, as Andy said, is to try one
and see what you get.

However, let me add another 2 cents worth here:  when planning library
capacity, the mistake people usually make is assuming that if they have 200
GB of data to back up, they need 200 GB of tape capacity in the library.
YOU WILL NEED MORE THAN THAT.

Over time your storage pool tapes will become fragmented as old versions of
data become invalid.  If your reclamation threshold is set at, say, 70%,
your tapes will all have different amounts of VALID data between 100 and
30%.  So you will need enough library capacity to account for the fact that
not all tapes are 100% full.

***************************************************************
Wanda Prather
The Johns Hopkins Applied Physics Lab
443-778-8769
wanda_prather AT jhuapl DOT edu

"Intelligence has much less practical application than you'd think" -
Scott Adams/Dilbert
***************************************************************




> -----Original Message-----
> From: Andy Raibeck [SMTP:storman AT US.IBM DOT COM]
> Sent: Wednesday, November 18, 1998 1:46 PM
> To:   ADSM-L AT VM.MARIST DOT EDU
> Subject:      Re: Client Compression Rate
>
> Hello Linda,
>
> It is almost impossible for me to tell you what the ADSM compression
> ratios would be for your data. This is because compression ratios
> vary, depending on the type of data being compressed. For instance,
> text files compress very nicely while files that are already
> compressed will not compress very well; in fact, they may even grow!
>
> Even your 3575 won't realize 3:1 compression on data that is already
> compressed. A very popular question goes something like, "IBM says
> I should be getting 30 GB on a 10 GB tape with 3590 compression, but
> my tapes are getting full and they have only 9 - 11 GB of data on
> them. What am I doing wrong?" The answer to this is that they are
> probably backing up a lot of data that is already compressed (like
> .ZIP files) or they are using client compression.
>
> ADSM uses an industry standard compression algorithm so if I had to
> give a number, I'd say 2:1 to 3:1, depending on the type of data
> you are backing up. Again, though, your mileage may vary!   :-)
>
> I would recommend backing up one or two systems that are
> representative of your clients and see what kind of compression
> ratios they yield. Then you can try to extrapolate from there.
>
> Sorry I can't be more accurate.
>
> Regards,
>
> Andy
>
> Andy Raibeck
> IBM Storage Systems Division
> ADSM Client Development
> e-mail: storman AT us.ibm DOT com
> "The only dumb question is the one that goes unasked."
>
> This is probably directed more to IBM but I would appreciated it if any
> one
> has any info or can point me in the right direction.  (I feel like I am
> going around in circle)
>
> I still have questions concerning the Client Compression Rate.
>
> Server: AIX 4.2.1 with ADSM 3.1.5
> Clients:  NT ver 2.x.x
>           NT ver 3.x.x
> Tape Library: 3575-L24
>
> I have client compress on due to network issues.  I need to est. the
> amount
> of tapes needed for roughly 600 PCs.   My question is what kind of
> compress
> rate can I expect using Client Compress.  (I am aware of the compression
> ratio for the 3575 - raw 3:1).  I placed a service call but so far all
> they
> have told me is the capacity of the library.  HELP! If I get an answer
> back
> from IBM, I will post it.  Thank in Advanced.
<Prev in Thread] Current Thread [Next in Thread>