1. 21
  1.  

  2. 2

    It’ll be interesting to see just how big the VMS on X86 market actually is.

    I rather wonder if it’ll mostly just be installations running atop aging DEC Alpha installations that’ll move over.

    1. 2

      I wasn’t even aware OpenVMS is available for x86_64. Though some legacy critical software designed for VMS “just works” after several decades, and there is less incentive to port it to newer architectures (ref. DEC Alpha). If I recall correctly, Indian Railways had their reservation system software running on VAX/VMS, and I won’t be surprised if they still continue to do so.

      1. 2

        It has been in the works for a long time now, this is them finally getting ready to deploy it. It’s still not production quality yet apparently and thy have been at it.. a decade at least.. it seems that long anyway! :)

        1. 3

          I don’t understand how one pays for such a long term engineering project. I mean if this was ones hobby, sure, put a decade into it. But for a commercial product? How does that make any sense? How is there a market for that after such a long time?

          1. 3

            They’ve continued to offer support on the older hardware during the transition. Remember VMS has been alive since the 1970’s and offers some pretty amazing features, like active-active failover and shared load. They were doing stuff like Erlang’s BEAM is capable of, but for entire Operating Systems, not just applications and have been delivering it in production for decades.

            So there is a very conservative user base that’s been using VMS for decades and decades. It’s very reliable tech at this point. This would be the 3rd or 4th architecture port for VMS.

            They have had demos of it running on X86 hardware for at least a year now.

            Our financial/back office system was on VMS when the Alpha hardware got killed and they announced the X86 port. We decided to just re-write the application using PyQT and Postgres. It’s been mostly painless for us, our uptime isn’t way less than VMS, but it’s adequate for our needs. We are at 99.7% uptime(24/7, including maintenance) even though we only have an 8-5 uptime promise(which we mostly succeed with) . With VMS 99.9…% is totally achievable with much less effort than we’ve put into our systems.

            It’s a different type of Computing, one the x86/Linux/Mac/Windows world is still mostly unable to match in capability. The IBM mainframe systems have been running identical software compatibility since the late 1960’s and are still running just fine today.

            1. 1

              They’ve continued to offer support on the older hardware during the transition. Remember VMS has been alive since the 1970’s and offers some pretty amazing features, like active-active failover and shared load. They were doing stuff like Erlang’s BEAM is capable of, but for entire Operating Systems, not just applications and have been delivering it in production for decades.

              I don’t doubt the quality of the system and its design, but there are so few users of it, one has to wonder how this all works out. No company will run something out of nostalgia. Having a single vendor for some niche is probably seen more as a risk than a benefit.

              So there is a very conservative user base that’s been using VMS for decades and decades. It’s very reliable tech at this point.

              People retire, move on etc. The person that introduced VMS at some company in 1985 or even 1995 is very likely no longer there. There must be some extremely critical stuff making substantial amounts of money, otherwise people would have moved on. That has nothing to do with the quality of the system, but more with the long term thinking. Again, I am not arguing that any newer system is better, just what I would think would make sense to me.

              Our financial/back office system was on VMS when the Alpha hardware got killed and they announced the X86 port. We decided to just re-write the application using PyQT and Postgres. It’s been mostly painless for us, our uptime isn’t way less than VMS, but it’s adequate for our needs. We are at 99.7% uptime(24/7, including maintenance) even though we only have an 8-5 uptime promise(which we mostly succeed with) . With VMS 99.9…% is totally achievable with much less effort than we’ve put into our systems.

              This I find a little funny in the positive sense. On the one hand we hear how super-crazy amazing VMS is, yet using PyQT and Postgres is just as good in delivering on the given SLA. It is probably also inifintely easier to find programmers for PyQT & Postgres, than VMS.

              1. 2

                I don’t doubt the quality of the system and its design, but there are so few users of it, one has to wonder how this all works out. No company will run something out of nostalgia. Having a single vendor for some niche is probably seen more as a risk than a benefit.

                Don’t mistake “I don’t hear much about VMS” for “There are few users of VMS”. It has a surprisingly robust community. Is it as big as Windows or Mac or Linux? Not by a long shot but there’s both interest AND money there for the folks who choose to occupy this niche.

                1. 1

                  There must be some extremely critical stuff making substantial amounts of money, otherwise people would have moved on.

                  Well, yeah?

                  job^’s ERP system was written in COBOL against a System/360. There have been ~3 attempts to “modernize”, each of which failed at various steps – some after lots of investment. This is a tale as old as time.

                  It’s remarkably difficult to move half a century of business rules and processes to a new environment. It’s even harder when that environment works well enough now and powers 9+ figures of revenue.

                  job^’s ERP system now runs on a leased System Z and is fully managed – turns out there is a very healthy ecosystem of MSPs that will run your mainframe, modernize and build new logic into your legacy processes, etc.

                  1. 1

                    sure, but that sounds like IBM Mainframe, not VMS. How does the success of the one speak for the other?

                    1. 2

                      You’re right to be confused; I managed to quite inarticulately lead with the example and not focus on the key point I was trying to make. Long day :)

                      In my experience, these legacy platforms live not because they offer something uniquely compelling. Instead, they survive because they’ve accreted often implicit knowledge/rules/business processes over a very long time. And that makes moving away from them more dangerous than sticking with them, as long as you have people you can pay to maintain them and a source of replacement parts – or enough spares in a closet.

                  2. 1

                    This I find a little funny in the positive sense. On the one hand we hear how super-crazy amazing VMS is, yet using PyQT and Postgres is just as good in delivering on the given SLA. It is probably also inifintely easier to find programmers for PyQT & Postgres, than VMS.

                    Agreed, a little funny! But we didn’t need anything VMS was really offering, I mean we took advantage of it, since we had it, but our SLA is 8-5 business hours basically. I mean Windows 95 can even manage that most of the time :)

                    As for finding programmers, agreed, ones with Python experience are easier to find, but in my estimation/experience, if you have a programmer that can’t learn a language and get useful quickly, they probably aren’t worth hiring, so language experience is mostly a pointless metric. That said, you generally want at least 1 person with a good level of experience in whatever language you use, so you can avoid the big language pitfalls newbies might run into not knowing any better and get pointed to the saner libraries for solving problem Y with less research required.

                    We mostly ported because users wanted “GUI”. Not that for most accounting/back office, a GUI necessarily helps, users still mostly want row/column output and reports yesterday that they dream up tomorrow. All easily done with a CLI/TUI.

                    1. 1

                      As for finding programmers, agreed, ones with Python experience are easier to find, but in my estimation/experience, if you have a programmer that can’t learn a language and get useful quickly, they probably aren’t worth hiring, so language experience is mostly a pointless metric. That said, you generally want at least 1 person with a good level of experience in whatever language you use, so you can avoid the big language pitfalls newbies might run into not knowing any better and get pointed to the saner libraries for solving problem Y with less research required.

                      Sure you can learn any language, but finding people who want to learn a completely different OS that is not much used anywhere else is probably not that easy. If you told me I had to leave all my Unix knowledge behind and work on some obscure system, I would not be interested. You will probably find people that want to do that, but the pool of people will def. be a lot smaller. I can learn Rust for a new job or Typescript or whatever, but if you make all my knowledge about the systems irrelevant, I will feel very uncomfortable and look elsewhere.

                      1. 3

                        We used Python on VMS also :)

                        Mostly a side note, VMS isn’t THAT different from *nix on a day to day level. I came from UNIX, did VMS for a while and never had any big trouble, and then went right back to FreeBSD & Linux hardly skipping a beat.

                        It helps that VMS documentation is generally fine, but it is a for pay product that takes some pretty big $$‘s to buy, so it’s not really surprising that their documentation existed and was decent. Unfortunately today’s world regularly punts on the documentation problem, and in some cases even punts on the customer support problem.

                2. 2

                  I think in part it isn’t purely commercial but also borne out of belief in VMS and wanting to preserve it

            2. 2

              I assume the Alpha sites have mostly moved to Itanium.

              1. 1

                It’s been a few years ago since I last heard from someone with a large VMS installation, but less than a decade ago the folks I talked to were still running a mix of hardware including VAX. Reliability is generally more important than performance to the folks still using VMS and some of the old VAX hardware is incredibly reliable (and was incredibly expensive when new), they only replace it with Alpha / Itanium stuff when it actually fails. Compaq was still selling VAX hardware until around the early 2000s. I think the last VAX systems were in roughly the same performance ballpark as a Pentium or Pentium II, so slower than a software emulation of one on a cheap mobile phone today (though some had quite powerful vector units for the era).

                1. 1

                  I’ve heard the opposite; there’s a lot of customers that stuck with Alpha because the Itanium boxes were too much for them.

                  From what I’ve heard, the people on real steel tend to be 1. midwestern auto parts shops that adopted VMS back when it was accessible to smaller businesses, and got stuck on it since 2. Fortune 500s. The old technical userbase mostly migrated.

                2. 1

                  From what I’ve heard, most people who still use VMS applications run them under emulation:

                  https://www.stromasys.com/solutions/charon-axp/

                3. 2

                  I wish I could just download an ISO and poke around without having to have a support contract.

                  1. 1

                    If they want to have a larger community of sysadmins, developers and users, this ought to be a high priority for them.

                    1. 1

                      VMS Software does have a community license program like the old hobbyist license program that HP ran, and it appears that they do plan to offer this for x86 as well, but they don’t currently. But if you’re endowed with sufficient reserves of patience, you could request a community license for the Alpha version and install it in an emulator.

                      1. 2

                        I don’t really want to register to download it, I want to just download it.