On Thu, 5 Sep 1996, Pittson, Timothy ,Corp,US wrote:

> 2) I've had good luck putting the database on a 7137 disk array (RAID-5)
> and the recovery log (mirrored) and disk storage pools on SSA disks.
> I'd definitely try to keep the database, recovery log, and disk storage
> pools on separate physical devices regardless of what kind of disk
> drives you use.
> 3) We did look at DLT but ultimately decided on 3590 drives.
> 4) I've been very satisfied with our 3494 and 2 3590 tape drives - we
> have the 2 tape drives hanging off of 2 separate, dedicated SCSI
> controllers.  Performance is excellent for everything except when
> migrating a lot of small files from disk storage pools to tape... this
> is more a limitation of ADSM than the hardware.  We're able to do a full
> database dump (8 GB) to a 3590 cartridge in under 40 minutes.

Your throughput is (8*1024)/(40*60) = 3.413 MB/sec. That is 1/3 of the
theoretical performance of the drives. To which degree is your database
filled? Have you seen faster values? We are considering buying 3590, so I
am curious about all informations I can get.

> 5) We're running ADSM on an RS/6000 R30 with 2 processors. We were
> running the 2.1.5.x code and the most CPU I saw ADSM using was 75-80% so
> it was definitely taking advantage of the multiple processors.  However,
> I had to back off the 2.1.5.x code and go with the 2.1.0.x line of code
> because of some software problems.  I hope to get back to the 2.1.5.x
> code when the next maintenance level becomes available

See we just switched to ADSM server 2.1.5.9 yesterday running on a C20
with AIX 4.1.4, I am also interested in what kind of problems have you
seen?

> .
> Tim Pittson
> tpittson AT himail.hcc DOT com
>
> >----------
> >From:  Peter Talajic[SMTP:pmt2f AT UOTTAWA DOT CA]
> >Sent:  Thursday, September 05, 1996 9:15 AM
> >To:    Multiple recipients of list ADSM-L
> >Subject:       ADSM Server Selection
> >
> >Hello everyone.
> >
> >At the University of Ottawa, we are in the process of selecting some new
> >equipment to run our ADSM server.  Currently, we have about 90-gig of data
> >stored in backups.  The current server configuration is:
> >
> >- RS/6000 (model 520H)
> >- 3 x 2-gig drives (fast and wide)
> >- 2 x 1-gig drives
> >- EXB-120 robot having 116 x 8mm tapes.
> >
> >The clients we currently backup are:
> >
> >- 20 unix servers
> >- 25 Novell servers
> >- 1 NT server
> >- 200 desktop computers, mostly Windows 3.1/95 and some Macintosh
> >
> >I am soliciting suggestions (especially from IBM staff) concerning the
> >following issues:
> >
> >1.  What number of tape/disk devices per SCSI controller do people suggest
> >one uses with the server, i.e., to maximize throughput?
> >2.  Are there any suggestions as to how we should organize our disk storage
> >pools and database?  By this I mean:  Are there any advantages in locating
> >storage pools and database components onto (A) the same physical disk (B)
> >their own physical disk, or (C) it doesn't matter
> >3.  Does anyone have information as to the internal structure of ADSM's
> >database?
> >3.  Does anyone make use of and/or have opinions concerning DLT drives?
> >4.  What do people think of MagStar or any other types of robotics?
> >5.  When running 2.5.1.x of the ADSM Server software on a SMP machine, does
> >it truly take advantage of the multi-processors?
> >
> >Thanks to everyone in advance.  Peter.
> >
>

______________________________________________________________________
Disclaimer: These are my views and not those of my employer.

                        Andreas Floeter
Phone: +49 40 74 37 - 48 50     FAX: +49 40 74 37 - 27 55 (or 47 60)
email: Andreas.Floeter AT airbus DOT de

                Daimler-Benz Aerospace Airbus GmbH
Kreetslag 10, 21129 Hamburg; Postfach 95 01 09, 21111 Hamburg, Germany