NetNews Usenet Archive 1992 #31/NN_1992_31.iso/spool/bit/listserv/ibmmain/2897

Text File | 1992-12-26 | 3.4 KB | 76 lines

Comments: Gated by [e-mail]
Path: sparky!uunet!paladin.american.edu!auvm!IS.RICE.EDU!DBOYES
X-Mailer: ELM [version 2.3 PL11]
Message-ID: <[e-mail]>
Newsgroups: bit.listserv.ibm-main
Date:         Fri, 25 Dec 1992 23:53:33 CST
Sender:       IBM Mainframe Discussion list <[e-mail]>
From:         David E Boyes <[e-mail]>
Subject:      Re: Logical Partitions/Physical Partitions
Comments: To: [e-mail]
In-Reply-To:  <[e-mail]>; from "Leonard D Woren" at Dec
              24, 92 6:16 pm
Lines: 61

> Ah, but suppose that you have a processor failure on a dyadic(etc)
> machine?  I don't think that VM recovers from that, and therefore the
> preferred guest bites it.

If you have 2 processors and have not dedicated processors to the
guest, the hardware signals the machine check to VM, and VM takes
the processor offline. Processing continues on the remaining
processor, and new work is dispatched on the remaining processor.
If you have multiple CPUs defined to the guest, the machine check
may/may not be reflected to the guest, depending on if real
processors were dedicated to the guest or not.

In a large configuration, where you have N>2 processors and you
have dedicated processors to a guest, the guest gets the same
machine check it gets on the real iron, and gets to deal with it
in the same way.

> Even if you believe Richard's point of view, which I don't, you still
> have to take into account that MVS/ESA is still at least 5 years ahead
> of VM/ESA in dynamic reconfiguration capabilities.

Go read the VM/ESA 1.2 announcements. I think you'll find a
number of improvements that bring VM/ESA to a much more favorable
position in this regard.

> If you run brand z
> disks like we do, when you get an error in a page dataset, if it's MVS
> you can just PAGEDEL that page dataset and fix the disk.  You have to
> shut down VM (and therefore all guests!) to fix it if it's on a VM
> pagespace.  If the error is in spool, JES2 can be told not to use that
> track.  VM has to be shutdown for repairs.

You're missing the distinction between dedicated devices and
simulated devices again. If the devices are dedicated to a guest,
MVS can do all the nifty stuff that you mention and VM never
knows it happened, because, hey, it's a dedicated device, not my
problem, right? right. If those types of errors happen on disks
that VM is directly using, then you're correct. VM/ESA 1.2
addresses this problem pretty well, though -- it adds the
capability to stop CP from further use of a particular volume for
any of the CP dedicated areas like page or spool and dynamically
bring such areas in and out of production w/o a disruptive IPL.

If you're going to run serious production guests under VM, you do
it with dedicated devices, not full pack minidisks. By doing so,
you retain the advantages of both systems in error recovery and
system management. You also do it with VM/XA or VM/ESA -- not HPO
or SP. Most of the things that the MVSers have pointed out were
true under HPO or SP, but have changed radically since VM/XA, and
particularly since the introduction of VM/ESA.

> MVS/ESA V4 supports
> dynamically changing the hardware configuration, so that you can add
> and remove devices without an IPL & POR.  Where is the VM support for
> this?

VM/ESA 1.2. It's not clear to me how MVS deals with changes that
require an IOCP change, though -- doesn't that still require a
disruptive IML?

> /Leonard