ADSM-L

Re: ADSM Experiences

1995-09-05 01:12:47
Subject: Re: ADSM Experiences
From: "Keith A. Crabb" <KEITH AT UHUPVM1.UH DOT EDU>
Date: Tue, 5 Sep 1995 00:12:47 CDT
On Fri, 1 Sep 1995 14:38:22 EDT Paul Zarnowski said:
>We don't have anywhere's near 1.4T of data, but we do have 250 machines.
>I know our avg filesize is much smaller than 40Kb, so the number of files
>we manage in our ADSM database is correspondingly higher.  I think we
>have about 3 million files in our db now, and about 18.5 million database
>records.  We run on an RS/6000, 2 channels of disks.
>I suspect it is important to spread your db on as many spindles as you can,
>to optimize performance.  We haven't done that yet.  We have 2 duplexed dbvols
>on 4 different drives and 2 different channels.  We do see a bit of I/O wait
>(as reported by the unix sar/sag commands).  I'd be interested in hearing
>what others think about whether faster disks would help aleviate I/O wait
>or not.  I suspect much of the wait is due to seeking, and not so much to
>actually transferring the data, so I'm not sure that faster disks would
>help.  We're currently using Seagate Barracuda's now.
>
>If you think you'd find it useful to talk, feel free to e-mail or call me.
>..Paul
>
>Paul Zarnowski                     Phone:   607/255-4757
>Cornell Information Technologies   Fax:     607/255-6523
>Cornell University                 US Mail: 315 CCC, Ithaca, NY 14853-2601

Has anyone considered the usage of RAID disk arrays for the ADSM database
volumes?  It would seem that the higher I/O rates possible from the RAID
array would make an them an appealing prospect as well as the fact that
some array controllers now come with a variable amount of disk cache memory.
While the cache would probably prove useless for read data, a small amount
of write cache space could prove itself to be very valuable when a few
dozen clients are all running at the same time.

The ability to dynamically rebuild data from a failed volume also strikes
me as being useful as it could be used to avoid having to keep the database
copy volumes (although you do suffer downtime while the volume is being
rebuilt which you wouldn't suffer by using database copy volumes).

---
Keith A. Crabb         Keith AT UH DOT EDU
Keith A. Crabb         Keith AT UH DOT EDU
University of Houston  Operating Systems Specialist +1-713-743-1530
<Prev in Thread] Current Thread [Next in Thread>