ADSM-L

Re: Number of files per hour??

1996-08-12 13:47:16
Subject: Re: Number of files per hour??
From: Paul Zarnowski <VKM AT CORNELLC.CIT.CORNELL DOT EDU>
Date: Mon, 12 Aug 1996 13:47:16 EDT
Michael,

We are considering the same thing, but on a larger scale - probably 1 million
files per day.  We have found that our current performance bottleneck is with
ADSM's ability to update its database.  The smaller the objects you are
moving, the more of a problem this becomes.  I would love to see some good
performance studies showing the best way to enhance DB update performance.
I simply have not had the time to work on this here.  I know that IBM used
to do some of this type of performance work from the Washington Systems
Center.  I don't know if they still do this or not.

Frankly, I'm not sure that ADSM is up to doing this.  We might be able to
back up our mail servers, but in doing so I suspect we would eat up (or more)
the capacity of our ADSM server.

..Paul
--
On Mon, 12 Aug 1996 10:41:09 -0400 Michael Ryan said:
On Mon, 12 Aug 1996 10:41:09 -0400 Michael Ryan said:
>     I have several mail systems, each having approx. 4 million files. On
>     an average day approximately 10-20% of the files change, so I am
>     looking at incrementals that will backup approx. 400,000 - 800,000
>     files each day. The files are small, around 1k-2k on average, and my
>     initial tests indicate that I can get approx 20,000 files per hour.
>     The testing was done between a "HP-UX A.09.04 E 9000/847" and a IBM
>     ES9000. The testing does not appear to be touching the network, nor is
>     it taxing the mainframe or HP box. I would like to be able to backup
>     each of system within a 6-8 hour window.
>
>     Does any one have a similar situation, and if so how have you tuned
>     the system to move this many files?
>
>     Does any one have any firm numbers on the quatity of files that can be
>     backed up in a 1 hour period?
>
>     Any suggestions????????
>
>
>
>     ****Please not that the numbers above are very rough and only a result
>     of some prelimiary testing.
>
>
>     Thanks in advance.
>     Michael Ryan
<Prev in Thread] Current Thread [Next in Thread>