5 ms·

OpenVMS 9.2 for x86 will be released tomorrow (2022-07-14), so exciting

21 points by jandeboevrie 4y ago

usr1106 4y ago
I liked VMS in the 1980s. I have an account for it at the Living Computer Museum, and occasionally I play around a bit. But why would I seriously want to use that in 2022? What does it do better than Linux?
- IntelMiner 4y ago
  I would assume it's just propping up legacy codebases that are too expensive/difficult to be re-written Why have a solution when there's good money to be made kicking the problem down the road and all that
- Kon-Peki 4y ago
  > What does it do better than Linux? The job control and batch queues built into the OS allow for some really powerful workflows and I can't think of a way to replicate that in Linux without a lot of extra software that needs to be maintained. Source code management and file versioning built into the OS and all the triggers and customization you can do would enable some really interesting CI/CD without a need for any extra software. But I still wouldn't procure a brand-new OpenVMS server and start a business with it.
  - josephcsible 4y ago
    What's wrong with installing software?
    - trelliscoded 4y ago
      Anyone who's worked with z/OS or OpenVMS's queuing and distributed execution facilities will tell you that there's no comparable open source solution, even if you start installing 3rd party software. The usual solution is to go buy a NetApp Filer to use NFS's distributed locking and storage, but NFS doesn't understand execution queues and there's no easy way to do things like buffer output to local storage temporarily if the cluster splits. Filers are also very expensive, whereas distributed shared storage is built into OpenVMS at the filesystem level.
- trelliscoded 4y ago
  It's an enterprise operating system. Nearly any feature you can think of that you would want in a large business is already built into the operating system. Clustering, applications built out of objects written in different languages, transparent network file support, automatic file versioning, backups, huge data sets, pervasive runtime performance and process introspection features, layered security, disaster recovery file set shadowing across the WAN, built in bdb2 like database, it's all there. Unlike a typical Linux deployment, these operating system components were designed and tested as a whole, so the system tends to have little if any issues when dealing with unusual circumstances. Enterprise deployments for things like factories, nuclear reactors, water plants, live cable TV stations, and 911 call centers require a level of stability that you can't get with things like RHEL or Windows. I don't mean reliability, I mean stability. The Linux kernel is reliable, OpenVMS is reliable, but a typical Linux distribution is made up of a hundred components constantly churning on different release schedules, whereas it's normal to take a 20 year old Alpha binary written in 3 different languages and expect it to run on Itanium or x64 using an OS that's 2 major versions ahead. Just like z/OS or NonStop, it's entirely reasonable to have deployments which never patch the underlying operating system, and even ones that do patch regularly do so on a far less frequent basis, with a well-defined procedure to canary the changes to avoid taking down production.

21 points by jandeboevrie 4y ago

I liked VMS in the 1980s. I have an account for it at the Living Computer Museum, and occasionally I play around a bit. But why would I seriously want to use that in 2022? What does it do better than Linux?

IntelMiner 4y ago
I would assume it's just propping up legacy codebases that are too expensive/difficult to be re-written Why have a solution when there's good money to be made kicking the problem down the road and all that
Kon-Peki 4y ago
> What does it do better than Linux? The job control and batch queues built into the OS allow for some really powerful workflows and I can't think of a way to replicate that in Linux without a lot of extra software that needs to be maintained. Source code management and file versioning built into the OS and all the triggers and customization you can do would enable some really interesting CI/CD without a need for any extra software. But I still wouldn't procure a brand-new OpenVMS server and start a business with it.
- josephcsible 4y ago
  What's wrong with installing software?
  - trelliscoded 4y ago
    Anyone who's worked with z/OS or OpenVMS's queuing and distributed execution facilities will tell you that there's no comparable open source solution, even if you start installing 3rd party software. The usual solution is to go buy a NetApp Filer to use NFS's distributed locking and storage, but NFS doesn't understand execution queues and there's no easy way to do things like buffer output to local storage temporarily if the cluster splits. Filers are also very expensive, whereas distributed shared storage is built into OpenVMS at the filesystem level.
trelliscoded 4y ago
It's an enterprise operating system. Nearly any feature you can think of that you would want in a large business is already built into the operating system. Clustering, applications built out of objects written in different languages, transparent network file support, automatic file versioning, backups, huge data sets, pervasive runtime performance and process introspection features, layered security, disaster recovery file set shadowing across the WAN, built in bdb2 like database, it's all there. Unlike a typical Linux deployment, these operating system components were designed and tested as a whole, so the system tends to have little if any issues when dealing with unusual circumstances. Enterprise deployments for things like factories, nuclear reactors, water plants, live cable TV stations, and 911 call centers require a level of stability that you can't get with things like RHEL or Windows. I don't mean reliability, I mean stability. The Linux kernel is reliable, OpenVMS is reliable, but a typical Linux distribution is made up of a hundred components constantly churning on different release schedules, whereas it's normal to take a 20 year old Alpha binary written in 3 different languages and expect it to run on Itanium or x64 using an OS that's 2 major versions ahead. Just like z/OS or NonStop, it's entirely reasonable to have deployments which never patch the underlying operating system, and even ones that do patch regularly do so on a far less frequent basis, with a well-defined procedure to canary the changes to avoid taking down production.