[ADSM-L] Who/What determines tape compression

2007-04-30 13:13:26
Subject: [ADSM-L] Who/What determines tape compression
From: Zoltan Forray/AC/VCU <zforray AT VCU DOT EDU>
Date: Mon, 30 Apr 2007 13:13:10 -0400
TSM servers - AIX v5.3.4.1        AIX 5.1 (yes I
know...unsupported....working to phase out and replace with Linux

I am currently dealing with a strange issue relating to data compression
levels (or the lack there-of) I am hoping someone can explain or point-out
where to look.

We originally had 4-3592E05 drives in our 3494 tape library. We just added
3-more. FWIW, these drives have the new "encryption" feature (we didn't
ask for it.....were told we had to take it.....cost us $3,000 more).

Ever since we added the new drives, I am noticing a drop in the
compression I am achieving from these drives on the AIX TSM server that
owns the library.

Before these new drives were installed, I achieved 1.8TB on one cartridge
owned by this server. I didn't do much 3592 usage since I simply didn't
have the drives to share amongst 4-servers.

Since these new drives have been put into use and I have agressively
started migrating data from 3590K tapes, I am only seeing <600GB per
cartridge. This is almost no compression at all.

I did a complete removal and re-add of all AIX tape and TSM devices and
bounced the TSM server.  It is obviously seeing and using them, with no
issues. All drives have the latest firmware, D3I1_A38.

Nothing has changed with the DEVCLASS.  All 3592's are using the same
devclass. It is configured as 3592-2C.  Scaled at 100%.

No, I can't say that the lower compression is coming from the new drives,

Any thoughts on what gives ?  Why the change in compression ratio (this is
a high-mix of various kinds of data. I should be getting at least 1TB vs
my Notes systems which can't compress much and I get around 700-800GB per

Zoltan Forray
Virginia Commonwealth University
Office of Technology Services
University Computing Center
e-mail: zforray AT vcu DOT edu
voice: 804-828-4807

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