[net.unix-wizards] VM/370 security and performance

lcc.barry@UCLA-LOCUS.ARPA (12/18/84)

I've read several interesting criticisms of VM/370 and at least one uninformed
one.  There are  several good points about VM/370:

1) CMS is the nearest thing to a friendly OS IBM has come up with.

2) The entire microfiche for VM/370 (CP, CMS, IPCS, and RSCS) is about 6" thick.
OS/360 took up two card boxes; I assume MVS is even bigger.

3) Maybe some system programmers in u-wiz ENJOY working between 2 and 6 AM
when there's a new version of the system to build, test, and install.  With
VM, you do your work during the day in parallel with production use of your
machine.  When the system works, you come in ONE night to put it in the
production machine and make sure it works there.

4) The layered approach of VM/370 (370 apparency to virtual machines, Op
Sys in vm to present a favorite interface to users) keeps the interfaces
simple while allowing the users to have a powerful operating system.

Much of the simplicity and power of VM/370 comes from something it has in common
with unix: it was built by a few people (~4) for their own convenience and 
somehow got out.

Unfortunately, VM/370 has gotten more complicated with the addition of many
new facilities: 
	For control of virtual consoles: SNA virtual console communication 
service; message system service; single console image; logical device support
(this last is used by PassThru, VM's equivalent of TELNET).
	For communication between virtual machines:  Virtual Machine 
Communications Facility (VMCF) and Inter-User Communication Vehicle (IUCV).
	CMS/DOS is almost as big as the rest of CMS.
	Likewise for CMS VSAM.

In spite of its growing pains, VM/370 & CMS is still the cleanest, easiest to
use operating system for 370-like machines around.  (Unless you can talk
IBM out of VMIX, a unix built on a modified VM/370.)

barry

muth@amdahl.UUCP (John A. Muth) (12/18/84)

<<>>

In <6623@brl-tgr.ARPA>, lcc.barry@UCLA-LOCUS.ARPA writes:

> I've read several interesting criticisms of VM/370 and at least one uninformed
> one.  There are  several good points about VM/370:
> 
> 1) CMS is the nearest thing to a friendly OS IBM has come up with.

True, but that's not saying much. CMS is only slightly more useable
than CP/M. I have never been impressed by CMS.

> 3) Maybe some system programmers in u-wiz ENJOY working between 2 and 6 AM
> when there's a new version of the system to build, test, and install.  With
> VM, you do your work during the day in parallel with production use of your
> machine.  When the system works, you come in ONE night to put it in the
> production machine and make sure it works there.

Exactly right.

I've been with Amdahl for 4 years as a UTS systems programmer. I think
that VM/CP is the slickest thing I have ever seen. It allows me to bring
up test versions of UTS at any time without effecting anyone else. No
one should ever attempt operating systems development with anything
less.

(UTS is Amdahl's port of UNIX to 370 compatable machines)

> In spite of its growing pains, VM/370 & CMS is still the cleanest, easiest to
> use operating system for 370-like machines around.  (Unless you can talk
> IBM out of VMIX, a unix built on a modified VM/370.)
> 
> barry

Or you can call Amdahl and get UTS right now (and you don't have to modify
your VM to run it). We're very proud of UTS around here.

If anyone is interested in UTS, contact:

	Cyndy Ainsworth, Mail Stop 208
	Amdahl Corporation
	1250 E. Arques Ave.
	P.O Box 3470
	Sunnyvale, Ca. 94088-3470

Or call her at:
	(408) 746-8945

-- 
				John Muth
				(408) 746-6069
				...{nsc,hplabs,ihnp4}!amdahl!muth

[The opinions expressed are those of the author and not necessarily
those of Amdahl Corporation, its management, or employees.]

rcd@opus.UUCP (12/19/84)

> 3) Maybe some system programmers in u-wiz ENJOY working between 2 and 6 AM
> when there's a new version of the system to build, test, and install.  With
> VM, you do your work during the day in parallel with production use of your
> machine.  When the system works, you come in ONE night to put it in the
> production machine and make sure it works there.

True enough for the systems which run UNDER VM.  However, as barry pointed
out later in the same article, the once-simple VM has grown an unfortunate
number of features and enhancements.  SO, this means that you get to come
in between 2 and 6 AM to work on VM.

Don't get me wrong--there's still orders of magnitude difference in size
between VM and the systems that run under it, such as MVS.  You DO get some
of the same benefits of a small kernel that you get (?got?) with UNIX.
-- 
Dick Dunn	{hao,ucbvax,allegra}!nbires!rcd		(303)444-5710 x3086
   ...Are you making this up as you go along?

bsa@ncoast.UUCP (12/25/84)

There is one feature of CMS that I would like to see under Unix: the
ability to type a key and have the current process move into the background.
Under Unix, it'd be nice if stdin were immediately redirected to /dev/null
and standard output to ./prog.out or something similar.  This would come
in handy when root writes to me in the middle of a long compile...

--bsa
(I don't know if 4.2 or SysV has the ability to do something similar
if it's written into each separate program -- but a kernel-level method
would be better than modifying and recompiling everything.)
-- 
  Brandon Allbery @ decvax!cwruecmp!ncoast!bsa (..ncoast!tdi1!bsa business)
	6504 Chestnut Road, Independence, Ohio 44131   (216) 524-1416
    Who said you had to be (a) a poor programmer or (b) a security hazard
			     to be a hacker?

josh@v1.UUCP (Josh Knight) (12/27/84)

> There is one feature of CMS that I would like to see under Unix: the
> ability to type a key and have the current process move into the background.

Have you heard of ASCII control-z (032 (octal), I think)?  Been in C-shell
since early Berkeley VM Unix.  Don't know about System N.

Could you point me to the "background" feature you speak of for CMS?  Are
you speaking of "CMS Subset", or what?

The views, confusion and anything else herein are not necessarily those of
my employer.


			Josh Knight, IBM T.J. Watson Research
    josh at YKTVMX on BITNET, josh.yktvmx.ibm on CSnet, ...!philabs!v1!josh

jbn@wdl1.UUCP (01/16/85)

     No, that's the beauty of VM; VM WILL RUN UNDER VM!  It is actually
possible, strange though it may sound to the UNIX community, to run a
full-fledged operating system, interrupt handlers, device drivers,
page table updating and all, UNDER THE CONTROL OF THE OPERATING SYSTEM.
There's a substantial performance penalty, but it's on the order of 30%,
not 300%.  So operating system development takes place along with other work.
     This technology was first developed for the IBM 360/67; a certain
amount of special hardware support is needed to make it work, but less than
one might think.  The key idea is that it must be possible to simulate
correctly all privileged operations.  Incidentally, operating systems of this 
type are called ``hypervisors''.  Few kinds of CPUs are capable of supporting a
hypervisor, but all 370s, 43xxs, and 303xs can do so.
     And yes, you can run VM under VM under VM ..., losing 30% performance
with each layer.

jbn@wdl1.UUCP (01/18/85)

     A cursory reading of the M68000 manual indicates that them M68010 and
M68020 are capable of supporting hypervisors.  How about a virtual machine
system for the M680x0 line?