ADSM-L

Re: Concerns

1994-10-27 13:22:31
Subject: Re: Concerns
From: "Wayne T. Smith" <WTS AT CAPSLAN.CAPS.MAINE DOT EDU>
Date: Thu, 27 Oct 1994 12:22:31 EST
Excerpts from Katherine Faella's article...
> I'm also concerned about the database.  How large will it grow?

The *minimum* is about 800 bytes per workstation file backed up.
So, a 20,000 file RS/6000 requires 20000*800=16,000,000=16Meg of
data base storage.  Now you don't want a full data base (full data
bases cause *lots* of problems!), so maybe 20M is required.  But
files change over time.  If you backup everything including the
operating system, then I'd estimate 10-20% more is required for
changed files.  Depending on how many versions you keep and if you
allow archive, the number can go higher.  If you mirror your data
base, double the result.

The requirements will vary wildly from client to client.  We have
AIX systems with 12,000 files and others with 250,000+ files.  On
those same machines, some change 1Meg nightly, others change 1Gig
nightly.

I wish (and my operators wish) we had a tape robot on our VM server!

Probably not worth much, but we find our Mac clients have a few
thousand files, our Windows and OS/2 clients have several thousand
files, and Netware servers can have anything.

 > How many people are going to be needed to go about campus helping
 > users install clients, setup their options etc?

I'd like to hear other's experiences in this area.  I've found I
should be able to install most clients in 30 minutes or less, but
it's never been done in less than 2 hours yet.  If they do it on
their own, the manual is too big to read.  I wrote two pages of
directions: *invariably*, they skipped some step ... one person
skipped exactly every other step!  :-)

We've also found that you can't assume that continuing backups are
being performed, because every client will do that for you :-)
Things break.   We started by FTPing/LPRing the schedule log to a
host for processing, but have found that this process breaks more
often than ADSM. If/when we go outside our computing department staff
machines, we will periodically process information available at our
(VM) server (and send mail when we suspect a problem ... and
sometimes just summarize via mail).  It's just too costly for us
otherwise.

> And what about when a disk fails?

ADSM is a lousy system if a disk fails.  Not only can't you just
start restoring (if some part of the operating system is affected),
but chances are you've got a lot of data at the other end of a
relatively slow pipe.  It's *much* better than nothing, but IMHO
having a disk fail will cost you and your client a lot of time.

>  If the average disk fails once every 4 years ...

No way.  Disks are much more reliable than that.  But it's still a
potentially big problem/expense to get these people back up and
running.  And maybe because you have agreed to backup their data they
will expect you to help in getting it back (such as restoring the
operating system, the communication software, tailoring the
communication software, installing ADSM and tailoring ADSM) ... could
be a lot of work.

Then lets say you've restored everything except the ADSM backed
up data ... and are tired and decide to complete things tomorrow (or
the failed volume does not cause system failure).  You've formatted
the new volume, but not populated it yet.  ADSM comes along and backs
up the system ... sees the empty volume/filesystem and ... voila! ...
all the files on it are marked deleted/inactive.  Now you have a
major problem trying to decide that to restore.  You want to restore
to the state the volume was in a day or three ago, but without a lot
of work (and the schedule log), your only choice is restore
everything ever backed up.  I can almost guarantee that won't all fit
on your new volume!

cheers from Maine,

Wayne T. Smith
Systems Group -- CAPS        internet: wts AT maine.maine DOT edu
University of Maine System   BITNET/CREN: WTS@MAINE
<Prev in Thread] Current Thread [Next in Thread>
  • Concerns, Katherine Faella
    • Re: Concerns, Wayne T. Smith <=