How many VMS systems exist

[Uwe Zessin]

Ah, yes, those detached process tricks.

We have a customer who uses a software where all 'users' run in a single process. It ran on VAX and was ported to Alpha. It has already survived a 'Windows attack' - a company wanted to sell them a Windows solution, but it, well, let's say it failed ;-)

We now attempt to replace its RA3000 with a SAN solution and perhaps introduce an enhancement to span two computer rooms (we are talking about two DS20E, so 'data center' is not appopriate, I think).

[Martin P.J. Zinser]

Hello Uwe,

at least in our case we are not talking "detached process" tricks, rather three (to four) tier client/server. So there are really no interactive/network logins on the backends besides of the account needed to run the server apps.

Greetings, martin

[Uwe Zessin]

Last time I checked Oracle had the right license terms to deal with that, too ;-)

[Mike Naime]

Uwe:

Like Martin, my Backend servers are running the Oracle DB's that the Citrix Front-end hit whenever the App needs data.

There could be 600 processes running on the backend system with no interactive users!

I might have 500 concurrent users on a Citrix farm that is sending DB requests to a 1280 that only has a 16 user license on it. If it wasn't for the BMC Patrol monitoring processes, most days you would not see any users on the backend besides yourself doing admin stuff.

Mike

[Uwe Zessin]

Mike,
I can tell the difference between a user license, an interactive login (I even know that not all get counted against a license), a detached process, a UAF username and I think I know what an I/O multiplexer is ;-)

[Jan van den Ende]

Yes Uwe,

Some of our old apps use the same trick.
Way back in time they were ported from Prime, and the old documentation (still the most complete set for it!) even states explicitly that is made that way so that it only needs a 1-user license!.
But because of management reporting we have special programs active that (UGH!) scans the tracefiles to identify individuals using the app, so, if not in the terms of license, we DO count them as users. It shows that the 2 "license users" actually account for around 300 - 350 different users daily; about 80 - 120 concurrent.
btw, because of that history, it of course
allows only ONE process access to the collection of (Direct random access!) files that tries to carry the name of "database", and to add some "poor-mans" cluster-awareness to it, it now runs in Batch. It tries to find a copy of itself on the other nodes in the cluster, and if so, it $ SYNCHRONISEs to that. If nothing to synchronise upon, only then does the app itself get activated. (yes, there is a tiebreaker if 3 batches fall trough the sync together).

Jan

Jan

[Uwe Zessin]

Hello Jan,

as you correctly say this is not safe against race conditions. Have you thought about using a small .EXE that uses the lock manager to try and take out an EX-mode lock in supervisor or exec-mode? A nice thing is that the lock manager works cluster-wide and the lock sticks if it was granted in an inner-mode. If you run this with SYSLCK privilege, you can even synchronize across UIC groups.

Any process who comes late will have to wait (there is no big difference if you do a $SYNCHRONIZE or run another programm because the batch job is running anyway).


I am not sure I have understood the working fully, but would it help to have a autostart queue with a job-limit of 1? The autostart will make sure that the queue can fail over if the current node goes down and the job limit ensures that only one job is running at a time. The only downside I can see is that all jobs will run on the same node, but it will relieve the other nodes from processes that just wait.

[Martin P.J. Zinser]

Hi,

just to make that clear, we do not do this because of licenses. We have a couple of other systems where we do need many different interactive users (mainly in development) and they are appropriatly licensed. Also Oracle is not an issue for most of our systems since we use RMS with RU Journaling extensively.

Greetings, Martin

[Jan van den Ende]

Uwe,

Purchased software, no sources available to us.
And certainly no budget will become available for such (not MY opinion!) "trivial point":
1st. It has been runnning like this for 2 decades. "You are seeing ghosts."
2nd. The software is outdated anyway. Anytime soon the (Unix) replacement will make it obsolete. (Planned for mid-1997, so maybe about due).

Yes, I am aware of the potential race condition. That is why the "fall through" mechanism is just a little bit more complex, with checkbacks that check whether another instance is also trying to promote, then step back one step and wait randomly 2-10 half-seconds. It is a pourman's abuse of the locking by QUEMAN.
If nobody else is in or past the try-to-promote phase, only then proceed.
The idea of just a failoverable queue with joblim=1 has been our first choice when we wanted to automate failover, but we need to run multiple instances of this (separated by different UIC groups), and the quename is hardcoded into the EXE's.
Not the way I would have designed things, but remember, it is ANCIENT, PORTED software.

And the issue of evading the need of user licenses... we are running Unlimited User licenses on that cluster...

Jan

[Ian Miller.]

what I don't understand is if there is growth in the number of VMS systems in the world (and a corresponding bigger growth in the number of users served by VMS systems) then why are hp reluctant to say what this growth is. I know hp publish vms success stories but some figure to show vms is increasing not declining would help to counter the common misapprehension that vms is a decling platform.

[Jan van den Ende]

Well, maybe I should submit not a "success story", but a "reality story".

One of my son's former high-school class-mates went to the university to study IT.
At a given evening (just before his graduation) he came by once again, while I happened to receive a standby call (I have remote login).
Ofcourse he got interested, and after I solved whatever that problem was (must have been minor, can't remember) he got to asking what strange thing this was.
"This is the central proceesing of the backoffice applications of the Amsterdam Police" "looks strange, never saw the likes".
"Well, this is VMS" (and then, from a University IT graduate:) "NEVER HEARD OF"

The discussion that then followed ("five year uptime? Impossible, we would have heard of such systems") at least resulted in us agreeing:

IF things like that REALLY exist, then WHY is that info not available TO UNIVERSITY IT STUDENTS?.
In his words: "It could sometimes be really usefull if we KNEW about the possibility to have something like that!"

But we have to live with DEC/Compaq/HP marketing...


Jan

[Mike Naime]

Uwe: The App I work with now is a 3-tiered Application. Front-end, Middleware, backend (DB). It was written to use several hundred detached processes that get started up when you boot the system. Prior to this, I came from a shop where I had 30-60 interactive users on 3 3100 model 40's that I was used too monitoring for most of the 90's. It took me several months to remember that just because I did not see any interactive users doesn't mean that nobody is using the system.

Jan: Back when I was a student, you had the honeybucket, Vax, and other Mainframes that ran the University systems. I missed Card punches by one semester. :-) You knew about the "BIG" mainframes and what they did. Today the mainframes are missing from the universities. Everyone is learning about C/Java/Linux... The small stuff!

HP seems more concerned with selling printers/camera's/desktops to College students than promoting HP UX or VMS.

[Martin P.J. Zinser]

Hi,

well, that is one of my favorite soapboxes.

Back in the nineties (which is not that long ago!) when I was still doing nuclear physics, all of our on-line analysis and actually most of the off-line analysis was run on VMS systems. This was "mission critical" computing, don't laugh! Running a medium sized accelrator cost at that time between 20-30kDM (15k USD) plus there was about one experiment run for each student doing their PhD. People leaving the institute were well trained in VMS at that time and new how to use it in a real life environment. Currently the same institute is pretty much all Windows and Linux and certainly not on any hp gear (besides of printers). There are still a few VMS systems, but you do need a special reason to use them. Obviously new graduates know next to nothing about VMS. Loosing this account certainly has not hurt hps bottomline much, since the systems we used where to the best of my knowledge well discounted, but loosing mindshare and making it hard for commercial sites to find well trained VMS persons does hurt!!! Maybe the Itanium port would be a could opportunity to seed that market again?!

Greetings, Martin

[Ian Miller.]

parhaps this sort of thing may help
http://www.openvms.org/stories.php?story=04/04/22/7063765

[Antoniov.]

Well done Ian
Turin, city where I live, is orienting its business to new technologies.
Here it's born Torino Wireless project http://www.torinowireless.it/ to make a big wi-fi network.
I'm happy vms is Turin's Polytechnic, good news.

@Antoniov

[Willem Grooters]

Antonio,,

I like the remark:

Turin, city where I live, is orienting its business to new technologies.

GREAT ammo for Paul Jerrom (see http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=555089)!
But alas - my knowlegde of Italian is close to nothing (though I can deduct "leggi tutto" means something like "read all"). I would like to be able to read what's it about in a language I DO understand ;-) I think that hols for more people....)

Willem

[Antoniov.]

Willem,
I can't read your link :-(

@Antoniov

[Willem Grooters]

Antonio,

Don't know the reason (system unavailable ???), but easy enough. Scroll back in this forum, I think 2nd page, look for "VMS vs. Unix".

Willem

(Ian, no points for this one ;-))

[Willem Grooters]

Antonio (and others)

Found it.
Thi sis the right link:

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=555089

")!" was added to the link by ??what??

WG

[Manny DeAssis]

The URL shown below points to a recent ComputerWorld article on OpenVMS. It too mentions that magic 400,000 number. On a different note it also discusses VMS and it's longevity in a positive light (for a change).

http://www.computerworld.com/softwaretopics/software/story/0,10801,97032,00.html

Thought is was worth mentioning.

[Wim Van den Wyngaert]

But to my knowledge, VMS is in phase out in 90% of the companies ...

Wim

[Uwe Zessin]

That is not surprising when even the owner tells everybody to 'migrate to NT'.

[labadie_1]

I can understand that Sue/HP can't disclose the number of Vms systems, but I do not think this hidden number reflects the real number of Vms systems. I think nobody knows it, because the contracts are not up to date for a good number of customers, and some systems are bought with an OS, and anothe rOS is installed. I have often seen a bunch of Vaxes replaced by a powerful Alpha, so I am not convinced this number is very significant. I think the revenue from Vms licences is more pertinent.

[Ian Miller.]

(is this the thread that would not die? :-)

There where rumors of double digit growth in VMS sales this year.

[labadie_1]

Ian

I think you have some info from a good source. May I say that I think this increase was not helped (for France) by some agressive Vms marketing ?

:-(