I need to find out more information about the relative performance
characteristics of ADSM servers running on mainframes, both VM and MVS,
and on RS/6000s.  We currently run our ADSM server under VM on a 3090 400S
which is shared with a production MVS system (both are preferred guests
under a top-level VM).  Some are not pleased with the performance of
of the VM server and the question has been raised as to whether it would
be better to run the server under MVS, or on a dedicated RS/6000.  Can
some of you reading this list share first-hand experience in comparing
the performance of ADSM servers on different platforms, or point me to
an IBM white paper or ITSC red book that discusses this issue?

More specifically, some of the questions I am interested in are:

Are there any advantages to running the server under MVS instead of VM?
  The VM server doesn't take advantage of CMS multitasking as far as I know.
  Does the MVS server?  Does that make a difference?

How does the performance of the server on an RS/6000 compare with the VM
  server?  Some ADSM functions are fairly disk I/O intensive.  Can the
  RS/6000 server with SCSI disks handle the load as well as the VM server
  with 3390s?  What about 8mm cartridge drives vs. 3490 drives?  If we
  put ADSM on an RS/6000, would we need to get 3490E drives for it, or would
  8mm drives be fast enough?

What kind of data rates do you see on backups between clients and various
  servers on a 16Mb token ring with TCP/IP?  I realize that is heavily
  dependent on other factors such as network traffic, but does an RS/6000
  have an advantage here?  TCP/IP transactions (e.g. pings) seem to take
  much longer to the VM system than to an RS/6000.  Our VM system is
  connected to the LAN via a channel attached RS/6000, using the CLAW
  protocol.  Our VM system also has very heavy SMTP activity and up to
  165+ logon sessions via TELNET.

There is some thinking here that it would be better to run ADSM on a
dedicated RS/6000, but that would require a substantial investment in new
hardware, and we need to know whether the investment would pay off in
improved performance.  Any and all help will be greatly appreciated.

Thanks,

Larry Sanders                              [email protected]
University Computing Services              MAINT AT MIZZOU1.MISSOURI DOT EDU
University of Missouri - Columbia