[Info-vax] VMS - what is the current thinking amongst the user community

[Sector7]

On Mar 25, 9:19 pm, David J Dachtera <djesys... at spam.comcast.net>
wrote:
> Jon.Po... at sector7.com wrote:
>
> > It been a long time since I posted anything - Sector7 has been working
> > away at migrating various applications - not much change there - We
> > are releasing 3.3 of the migration libraries - which provide full
> > duplex asynchronous I/O, AST's (supervisor, kernel, user etc) - and
> > has tested at doing over 50,000 lock conversion per second on some old
> > Intel equipment.
>
> > So much for the bits and bytes - Obviously, the projects we see these
> > days are no longer the 'low hanging fruit' - we are - as always -
> > focused on complex code migrations - I would be very interested in
> > understanding what the VMS users "feel about" VMS. We seem to be
> > seeing large corporations with UNIX as the focus - that have a VMS
> > server / cluster that is working perfectly - but - represents a very
> > small percentage of the organizations skill set - and yet is as
> > mission critial as any other server.
>
> > We wont be soliciting / calling or even emailing - but - If anyone
> > feel so inclined I'd especially like to hear:
>
> > (a) Is your VMS system running mainly home grown application or
> > packaged
> > (b) If homegrown - are there any specific VMS API/ Subsystems - that
> > make VMS irreplaceable (lock manager, clustering etc)
> > (c) Is there any 'software' that would make life easier - I hate to
> > use the DEC I14Y "Interoperability" - only because - I still miss
> > DECUS / New Orleans and the TGV guys paying a for several great nights
> > out.
> > (d) Is there still the religious fervor associated with VMS? (after
> > having to duplicate the AST/QIO/Lock Manager mechanisms - I have even
> > a greater appreciation of VMS internals)
> > (e) What programming languages are popular ? (no one has asked for
> > BASIC or DIBOL for 2 or 3 years)
> > (f) If VMS was going to be replaced - as VMS guys - what target would
> > you lean toward (AIX/HPUX/LINUX/Solaris x86)
>
> > I appreciate anyone taking the time to read this.
>
> > Sincerely
>
> Hey! Gib Senip! Long time, no read! Looks like life at reknaw.com is
> going well...
>
> Y'know, there is one market niche I believe you are missing out on:
>
> AIX(, UX, Linux, etc.) needs something akin to VMScluster. HACMP is not
> "clustering", it's fail-over detection and initiation, at best.
>
> Since the healthcare industry is migrating away from VMS as you read
> this (HP went out and pushed UX on I64 instead of VMS; the ISVs read
> that as "VMS is EOL"), and many sites are moving their Cerner and
> InterSystems software to AIX, even the less technical Cerner, etc.
> people are finding the lack of VMS-style clustering to be the bane of
> their UN*Xland existence.
>
> One of the sorest points is the lack of any functionality approaching
> the "DO" command of VMS's SYSMAN utility. Action that used to be done
> once are now done repeatedly, consuimg vastly more time than before, and
> introducing lots and lots of opportunity for error.
>
> For example, we have 6 production LPARs and four in development. Done
> "the long way", a task as simple as adding a printer to the backend
> would take seven(7) (your favorite number?) separate printer adds, and
> that doesn't even account for having multiple development environments
> (add another nine(9) or so separate actions). (I've managed to work
> around that. I add printers to AIX once in Dev., once in Prod. and into
> Millennium once in prod. which is then propagated (queue.def and
> queue.cfg) to the other three prod LPARs and to the environment-specific
> paths on the three dev. LPARs that do printing.)
>
> So, if you actually have a functional DLM, the next step is the o.s.
> hooks to enable storage sharing they way we can in a VMScluster, though
> perhaps modifying "mount" to accept a "--cluster" command line option
> (or equiv.) may not be entirely practical in UN*Xland. Check it out...
>
> ...from MY perspective as a systems guy.
>
> D.J.D.

Hello David - yes - its been a long time - "Gib Senip / ReKnew" all
seems trivial now - but - I do miss the passion of the old days - our
lock manager is unfortunately not yet distributed - it was hard enough
- just getting unix to handle the various AST's and deliver them in a
code safe framework - in fact - its only been recent additions to the
SOME of the UNIX implementations of pthreads that has allowed a true
high speed 'safe' AST delivery mechanism. You are right about HACMP -
High Availabilty Clustered Multi Proessor - its not VMS Clustering -
The Oracle lock manager does provide some cluster wide resource
locking - but - requires code changes (and no AST's)

I'll pass your comments on - I thank you.