=> On Fri, 28 Jul 2000 11:41:05 -0400, Paul Zarnowski <vkm AT 
CORNELLC.CIT.CORNELL DOT EDU> said:

> Since there has been so much discussion on this, I'll add my 2 cents.  Here
> at Cornell, we have a NIM server which we use for doing AIX installs.  We
> run a weekly script which causes mksysb backups to be made for all of our
> AIX systems (about 30, I think).  This is all implemented using ADSM, NIM,
> mksysb, nfs, and 4 homegrown scripts.  Here's how it works:

[ ... succint description ... ]

Paul, you might want to make those scripts availiable; I anticipate that there
will be many more people interested.

I wonder if anyone has comments about keeping this (_HUGE_) filespace
hardware-attached to the ADSM server, so you can avoid dumping many multi-gig
files across a network when the mksysbs run?

We're beginning to experiment with NIM here at the University of Florida, too.
Your comments about NFS being inefficient will generate knowing nods all
around, but until we started playing with it, we didn't know just _how_
inefficient:

We did a bit of a comparison; a H50 and a B50, similarly network attached,
both install playgrounds at the moment (and thus having roughly equivalent
file loads). Similar processors and memory.

The B50 did a mksysb over a pipe to an encrypyed (!) SSH connection to the NIM
master.  The H50 did it over NFS.

The B50 was somewhat over _four times_ as fast.  I'm almost afraid of what it
might do if we recompiled SSH to permit a connection without encryption.

So, think hard about mksysbs to rsh's or ssh's rather than NFS partitions.


Allen S. Rout