SoylentNews

VMS Will be Ready to Run on x86 in 2019!

posted by on Thursday April 20 2017, @10:46AM   
from the breaking-news dept.

MrPlow writes:

Submitted via IRC for TheMightyBuzzard

Or 2018 if you're brave. For now, we have a boot screen!

Story's a bit dated but being as they're in no rush, I don't see any need for us to be either. So, you lot think we'll ever actually get to play with a VMS box on cheap hardware or is this going to be another DNF situation?

Source: https://www.theregister.co.uk/2016/10/13/openvms_moves_slowly_towards_x86/

Previous coverage:
OpenVMS Not Yet Dead.

---

Original Submission

• Re:Design comparison? Re:Design comparison? (Score: 4, Interesting) by mechanicjay on Friday April 21 2017, @04:22AM (1 child)

  by mechanicjay (7) <reversethis-{gro ... a} {yajcinahcem}> on Friday April 21 2017, @04:22AM (#497229) Homepage Journal

  Well, one of the things with VMS is it's really a mid-70's product, and a direct decendent of RSX-11 for the PDP-11 series of computers..and it shows. It was fairly well modernized through the 90's however, until DEC went tits-up. Then it became part of Compaq's portfolio, who, being a hardware company, had no idea what to do with it. Then quickly it became HP's product, where it competed against their own in-house proprietary OS's. So, it was basically left to languish by the year 2000. The port to Itanium never really went anywhere, as performance on Itainum, until the most recent iteration was basically barely on par with the old Alpha servers. Anyway, as to your design questions:

  • Virtualization: Nothing in the product, this is a bare metal OS. There are emulators out there to run it as a guest OS, but performance is complete shit.
  • Swap: VMS stands for Virtual Memory System -- kind of pioneered the idea. If you're familiar with the Windows Page File, it's basically the same thing.
  • Device abstraction: Between Symbols and Logicals (these two items could take paragraphs to explain -- part alias, part symlinks, all confusing) you can have so much abstraction you don't know which end is up. Combine that with way a single "directory" that can show the contents of two disparate sub directories (yes, on purpose) and ooof. Uhhh, so yeah, you can kind abstract stuff, not quite in the Unix way though.
  • Permission structure: The basic permission structure is quite limited. A user gets a UID and Group -- You can control access to files or executables on either, but there's no way to give a user multiple groups the way *nix does. There are extended ACLs, which you can apply to files/directories but it can get very complex very quickly and is really hard to audit, IMHO.
  • Administration: For a single machine, it's no better/worse than anything else. DCL is actually a very decent scripting language, especially considering it's vintage -- certainly more usable than any of the early Unix shells. I don't know about large scale cluster administration though, as I've never run more than a single node. I do know VMS clusters are known to have uptime measured in DECADES and clusters have survived rolling upgrades of all the nodes, so there must be something there. Also, I think for most places, their VMS cluster is kind of large monolith. This stuff comes from an era before commoditized x86 server proliferation, so I'm not sure you'd be trying to admin 500 VMS boxen -- its' just a different use case. That said, there's an opensource port of bash: GNV (Gnu's not VMS), and ssh is a part of all the TCP/IP packages, so there's no reason you couldn't do some sort of configuration management against VMS with a tool like Ansibel which just kind of fires commands over ssh.

  --
  My VMS box beat up your Windows box.

  Parent

  Starting Score:	1		point
  Moderation		+2
  Interesting=1, Informative=1, Total=2
  Extra 'Interesting' Modifier		0
  Karma-Bonus Modifier		+1
  Total Score:		4

• Re:Design comparison? (Score: 2) by kaszz on Friday April 21 2017, @06:55AM

  by kaszz (4211) on Friday April 21 2017, @06:55AM (#497274) Journal

  How is memory management? ie where a memory section in your process memory may be fragmented to just about anywhere in physical memory or even disc blocks.

  What makes it possible to migrate nodes? it must both transfer memory and instruction pointer, register setting, processor mode, device state etc and then there's multiprocessing. So it must more or less stop a processor, save all status to a communication line, verify and continue on another node. A lot of atomic operation issues and consistency. Though a modified scheduler on Unix should maybe be able to do the same?
  (I'm thinking this has some similarities to the "Freezer" cartridges in the past)

  Can a user mount their own filesystem, that only they can access? (like on Hurd) as for the rest it seems VMS more or less deals with the filesystem as a SQL database. Every storage object is a SQL row or similar paradigm.

  Parent