ADSM-L

Re: ADSM IO-Performance Tuning Tips welcome !!!

1997-10-30 16:50:15
Subject: Re: ADSM IO-Performance Tuning Tips welcome !!!
From: "Kauffman, Tom" <KauffmanT AT NIBCO DOT COM>
Date: Thu, 30 Oct 1997 16:50:15 -0500
We're running four DLT-7000 units in a Storage-Tek 9710 under ADSM
2.1.5.13. The server is a SP wide node (old style, 77 MHz) and the
clients are SP high nodes. I'm getting roughly 17 GB/hour direct to tape
for a single-tape session, and just over 30 GB/hour for a client running
to two drives -- but I'm talking large files here. My small files are 20
MB, the average size is closer to 450 MB.

And as near as I can tell, I should be able to run the drives faster but
I'm paging too heavily on the client side. (More tuning, I guess).

Oh - the drives are differential SCSI, and I have two per SCSI interface
in the server. I don't see any difference in performance when I'm using
one drive on each controller or both on the same controller for backups.
Both on the same controller seems a bit slower in doing tape-to-tape for
storage pool backups.

Tom Kauffman
Sr. Technical Advisor
NIBCO, Inc

>----------
>From:  Dirk Pohl[SMTP:Dirk_Pohl AT VISINFOSYS DOT COM]
>Sent:  Thursday, October 30, 1997 11:25 AM
>To:    ADSM-L AT VM.MARIST DOT EDU
>Subject:       ADSM IO-Performance Tuning Tips welcome !!!
>
>Hi all ADSM-ers
>
>We are installing ADSM on AIX
>
>Levels :  AIX       4.1.5
>          ADSM-Server   V2R1L5M13
>          ADSM Client   Level 7
>
>and we see, that ADSM doesn't use all the speed of hardware.
>We have an SSA-Subsystem and DLT7000 drives and we dont get the expected
>I/O rates with ADSM.
>Our DLT7000 drives perform with about 10-12 GB/hour with compression
>enabled
>(write speed is 5 MB/sec. w.o. compression so we should get at least 18
>GB/hour).
>The same effects we see with our SSA-Disks.
>They'll never perform over 4MB/sec and in middle with 2.5 -3.5 MB/sec with
>ADSM
>and no other I/Os are running on that machine.
>In an native environment(cp,dd,...) we get better device rating's with the
>same devices,
>so its our opinion,that it's a problem with ADSM.
>Maybe this is the way,how ADSM works , but I'm interested in other
>Benchmark Data
>with adequate devices .
>
>Has anybody tuning tips for me.
>
>
>regards
>D. Pohl
>(ADSM-Support)
>V.I.S. Informationssysteme GmbH
>Germany
>
>please respond to : dirk_pohl AT visinfosys DOT com
>               or to the adsm-l
>
<Prev in Thread] Current Thread [Next in Thread>