How small can VMS get ?

From: mist dragon (mistdragon_at_zdnetonebox.com)
Date: 05/16/03 

Date: 16 May 2003 01:55:55 -0700

Hi!

I read one post that told the trainer in HP telling VMS could become
'layered' product but since the subject was of totally other thing I
thought its better to create a new thread of the sobject.

There was a strong opinion that this is not the same as OpenVMS and
that is true. However, its a matter of what people envision as VMS. As
an Operating System or as a set of services ?

Does anyone remember how HP merged MPE ? In the beginning MPE was
ported to PA-RISC. In every new release these came closer and in the
end it became an emulator service running on top of HP-UX, finally
dying. What would prevent of HP redoing this with VMS when both HP-UX
and VMS will be ran on same hw platform ?

Let us analyze neutrally this possibility a bit.

The main drawback of emulation is its speed. The lowest level you
emulate, the slower is the emulation.

The lowest level is emulation of the processor architecture. Example
of this in DEC was FX!32 Emulation that was later expaneded with
profiler instructing compiling the least performing parts to native
code on the fly. While techncially sound, the performance never
exceeded the native performance and it never became a success.
Eventually this windows port was killed and FX!32 with it.

A similar, but simpler approach is taken in Charon-VAX that provides
the emulation of only hardware. VMS software is running on emulation
using real OpenVMS on top of it. This method is lighter than pure
processor level, but still very slow. The upside of this is that the
new pc's are faster than old VAXes so this does not really matter.

The lightest simulation would be to port VMS API to HP-UX, e.g.
provide a set of libraries for every compiler they use now for mapping
the calls from VMS calls to Unix calls because in most cases people
care only about their applications being supported, not of their
operating systems. In addition to this, they would need to create VMS
emulation shell for DCL.

Emulating VMS on Alpha was not reasonable on time the decision was
made because VMS user base was strong enough to sustain the porting
expenses, even thought the Alpha platform running Tru64 far
outpefrorms the same platform running OpenVMS on SMP with +5000 users
(disc io and ip communication suffocate).

Now both VMS and HP-UX and Linux will be on same platform. Looking
back to MPE this may be the future.

Naturally this could have interesting effects to performance because
mapping of API's is seldom 1:1. Otoh, if the performance of platform
running the emulation is good enough like in the case of Charon-Vax,
then only very few might care about that...

Relevant Pages

  • Re: How small can VMS get ?
    ... > Does anyone remember how HP merged MPE? ... MPE has never been 'merged' with HP-UX. ... MPE provides emulation for some 16-bit HP3000 stack architecture code ... its a matter of what people envision as VMS. ...
    (comp.os.vms)
  • Re: next VMS versions
    ... I'd say to be precise its a matter of what people envision as VMS. ... an Operating System or as a set of services. ... One view to services is FX!32 Emulation where the emulation is on ... OpenVMS is soon on same hw platform as HP-UX, ...
    (comp.os.vms)
  • Re: Could a PC do this?
    ... > essentially gives you a 'CPU', and the emulation is running on something ... then the realtime capability of VMS on the emulation should be as ... a robust OS capable of real time. ...
    (comp.os.vms)
  • Re: Anyone interested in building a vms-like OS?
    ... If it's just a DCL like shell - then the guys at ASCI (and I believe the ... If it means being able to take a VMS binary and run it, ... IMHO the best VMS "clone" is VMS running on top of a CPU emulation because ...
    (comp.os.vms)
  • Re: Setting Reflection 2 window titles from VMS
    ... emulation session from VMS. ... fix the titles. ... can have one script work for what ever is in use. ...
    (comp.os.vms)