Slashdot Mirror
HP Gives OpenVMS New Life and Path To X86 Port
dcblogs (1096431) writes Hewlett-Packard has changed its direction on OpenVMS. Instead of pushing its users off the system, it has licensed OpenVMS to a new firm that plans to develop ports to the latest Itanium chips and is promising eventual support for x86 processors. Last year, HP put OpenVMS on the path to extinction. It said it would not validate the operating system to its latest hardware or produce new versions of it. The move to license the OpenVMS source code to a new entity, VMS Software Inc. (VSI), amounts to a reversal of that earlier decision. VSI plans to validate the operating system on Intel's Itanium eight-core Poulson chips by early 2015, as well as support for HP hardware running the upcoming 'Kittson' chip. It will also develop an x86 port, although it isn't specifying a timeframe. And it plans to develop new versions of OpenVMS.
As a qualified Computer Systems Necromancer I've been disappointed by the lack of demand for combine technical aptitude with an ability to work with the undead creatures of nightmare. HP's plans are an exciting development for me and my colleagues!
I has them. I've been a huge fan of VMS since I first used, then later managed, DECstations at my university. Supported that platform for DECADES, and watched it finally go down the tubes under HP.
SO glad it's coming back!
The Vomit-making system returns from the dead in zombie form!
OpenVMS will outlive us all. I really can't believe there are that many OpenVMS boxes in the wild. Can anyone list some applications still being run by OpenVMS?
Only the State obtains its revenue by coercion. - Murray Rothbard
I would have thought every platform is decent for FORTRAN.
There are applications that VMS does very well in. Clustering under VMS is unsurpassed by anything else.
"To those who are overly cautious, everything is impossible. "
"I am surprised that people still want to use OpenVMS. "
Most Railways electronic Dispatching Systems run on OpenVMS, even the latest versions.
It's a bit like the Space Shuttle, you can't change anything without a crapload of red tape.
The only thing I don't understand in the article, is that they have been running it on x86 systems for 10 years.
OpenVMS is still used where high availability is needed but rarely at the front of a stack visible to users. Were I work, it's the back end, core application server (OpenVMS 8.4 on Integrity blades in a C7000 chassis), that without much effort stays up in the five or six 9's range, we use 2 or 3 CPUs worth of processing out of 16 and nobody complains about performance. Two of us easily survive 24/365 on-call because there is rarely a call. Changes from the software vendor or our in house programing staff are weekly if not daily, so it's not a static environment. We have no intent or desire to move to something else, there is little incentive: it would take 10 years to convert and certify, and several million dollars that we could us elsewhere (the study was done about 4 years ago when we moved from Alpha to Integrity). All that said, there is a lot of work needed to move to OpenVMS to X86-64: I don't expect anything for 5 years.
You joke, but "nightmare" would be an accurate description for today's youth if asked to work with VMS. :-)
We are talking about a CLI (DCL) which is so out of date you cannot even edit commands which span more than one line.
There's also no nice modern 1990s technologies such as filename completion as well.
The filesystem (ODS-2/ODS-5) is robust, if slow, however. It cannot handle upcoming multiple terabyte disk sizes however.
banks and insurance companies still use OpenVMS, which has clustering and filesystem features GNU/Linux and Unix have yet to evolve. Why do you mention Debian, it has no ability to run OpenVMS software
No there is no x86 code port, you might be thinking of emulator, which is fine if you want to emulate a microVAX workstation
As an old DECcie I loved VMS, but then I was coming from RSX-11m. It made for a pretty good software development system. One thing that can really hold it back though is its file system. While it is a robust FS, it is very closely coupled with the OS. There could be some real problems if/when people want to run it on something like ext4 or ReiserFS. Bodes for a whole new set of drivers and fs converters. Sigh.
Exactly: I'm sure there are tons of custom apps written for VMS in banks, insurance companies, railroads, etc. These are places where 'if it works, don't break it' rules, and VMS is working, and has worked for decades. Being able to buy support and replace hardware is valuable to them, and I wouldn't switch platforms in their place unless there was no other option.
'Sensible' is a curse word.
> I'm sure someone's crunched the numbers and this makes sense on paper, but seriously? Porting to Itanium before x86?
It is already ported to itanium, that happened years ago.
They are just talking about qualification testing of VMS on the latest itanium chips.
Are you confusing Itanium (IA-64) with x86 ?
Itanium is most certainly not x86. VMS runs native on the former; it does not run native on the latter.
VAX was already on 64-bit for ages when Linux was still in it's earliest versions. It's not going 'x86'. It's going 'x86-64', which didn't exist when Itanium was created. IA-64 was Intel's vision of the future - a complete overhaul of the instruction set. It bombed, but AMD64 wasn't written until several years later - and AMD does nice chips, but they don't really compete in that segment. (Or they didn't in 2001, at least.) It made perfect sense to port to what was supposed to be the new enterprise-class processor, instead of porting to an outdated desktop-class processor.
Linux on x86 can do lots of things, and is a very good system for many situations. If you need big iron (and the capabilities it provides - things like being able to upgrade or replace CPUs on running machines without downtime), VAX is better. In many cases you don't actually need big iron - a cluster of Linux boxes will do just fine. But when you need it, nothing else will do.
'Sensible' is a curse word.
I'm sure someone's crunched the numbers and this makes sense on paper, but seriously? Porting to Itanium before x86? I know HP wants to prop up its teensy niche CPU server line, but I just can't see how to justify that.
The reason is they hardly have to do any work for Itanium; they just have to QA a 8-core system instead of a 4-core one. The original port was done over a decade ago. With 20/20 hindsight it was a wrong move, the right one being presumably to tell Intel to shove it and wait a few years for the x64.
Who's going to migrate software from old VMS systems to a new one on very highly vendor-locked hardware?
Someone that has a 2 or 4 core processor Itanium system already. If anything is a non-starter it's the x86 version.
VAX was already on 64-bit for ages when Linux was still in it's earliest versions.
VAX -- the CPU architecture -- was always only 32-bit. If you mean VMS -- the operating system that was ported to 64-bit Alpha and then eventually to Itanium -- then we're good.
NT is no VMS. It wishes.
After UNIX/Linux, VMS is my favourite OS followed by RSX and TSX 11 from the PDP-11. I'm old, but DEC rules!
Because DEC/Compaq/HP never screwed it up by insisting that mundane software run with ring 0 privileges?
The sound of hands clapping by all zero remaining Itanic lusers.
I want to know how many legacy VMS users there really are left out there. It's been SO long that companies have been forced to start researching migrating off of VMS, and I expect that a lot have made the jump.
I'm wondering if this is more an attempt to bolster staggering Itanium sales than it is to really make VMS strong again.
Do not look into laser with remaining eye.
I think that the controller in my air conditioner is powerful enough for FORTRAN programs...
Do not look into laser with remaining eye.
no really, a decnet network for hobbiest, enthusiasts.
There are literally dozens of us!
Because they're not at all the same thing, they're not even close. There's only a superficial resemblance, very high level concepts only. Also the concepts they have in common are very often very common in many operating systems! I think the article was written by someone who'd only ever seen VMS, NT, and Unix and failed to realize just how much variety there really was out there.
(and Cutler was called in originally to do OS/2, which is also not like VMS)
Declare a platform dead one year, support it again the next year once customers had time to think about migrating away. Product strategist at HP seems to be a very nice job.
D'oh. Sorry, yeah, my bad fingers. VMS, not VAX.
'Sensible' is a curse word.
Exactly. NT3 was cool, NT4 was turned into Windows and hid the WNT foundation as much as possible. And put the graphics in ring 0, shudder.
thegodmovie.com - watch it
http://www.freevms.net/
DEC introduced some very handy and useful extensions to FORTRAN and had an awesome compiler, and these non-standard extensions ended up getting supported by non-DEC compilers. They also had an outstanding language reference manual. In my experience, almost everyone who used a VAX for programming did so in FORTRAN while almost all the C programmers did their work on the Sun workstations.
You are right, HELP was awesome.
Don't forget DECNET. That completely blew me away when I first used it. I accessed a file from half-way around the world just like it was on the computer right next to me. It was this wonderful interconnected web of VAXes all over the world that acted like a local cluster.
I've been using VMS non-stop for 22 years, starting in college (on a real VT-100 terminal in a lab full of them in the basement of my dorm; used VAXPHONE to talk to friends back home; VT640s were pretty neat too) to today, writing new Fortran applications and supporting legacy FORTRAN applications that have their roots in the late '60s. We have a mixture of "old" Alphas, EV68s that still rock and have uptimes measured in years, as well as some new Itaniums. The older I get, the more I am amazed at how very well designed and DOCUMENTED!!! VMS is compared to any other OS I've ever used -- just about all of them it seems.
Aproximately 1000 years ago, in a galaxy far far away, I learned my trade on VMS on an old Vax mainframe doing cobol. Horrible horrible stuff. But it was a rock solid operating system with features that you just don't see anymore, and more to the point having the code out there so coders can see another way of doing operating system far from the Unix or windows mainstream has a lot of value in and of itself.
Heck maybe people might port it around (Difficult job though. The old VMS had a .... unique...... way of doing things with its register marks and bizaro addressing modes) which could provide options for people who want to utilize VMSs ultra-secure design.
Excuse the Unicode crap in my posts. That's an apostrophe, and slashdot is busted.
Implementation makes a difference. Early versions of NT were quite good, but unpopular because you needed 16MB of RAM (if I recall correctly) to run them in an era when a high end personal computer shipped with 4MB of RAM. Over the years they tried to hold the line, at one point getting the minimum down to 12MB of RAM, but perhaps not coincidentally stability got really bad.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
OpenVMS is the most mature microkernel OS out there. You can have flaky hardware, flaky drivers, flaky software, and it'll just keep running perfectly, restarting whatever services as need, as often as needed. You can't make it panic.
It also has more advanced clustering than most people believe exists... A server's full state is replicated in real-time, so a hardware failure doesn't even need to be handled by applications, they just think everything has been running for the past decade...
OpenVMS has ridiculous uptimes, over a decade, even on heavily utilized systems. Far longer than anything else out there.
http://www.uptimes-project.org...
http://www.osnews.com/comments...
Slashdot gets worse every day... Pipedot: News for nerds, without the corporate slant
And when you want a SpaceX rocket, why not go buy a Tesla car instead?
Slashdot gets worse every day... Pipedot: News for nerds, without the corporate slant
The expectations of most programers and consumers have shifted to not consider uptime all that high a priority. It is generally only older shops in conservative industries (or conservative departments within, like payroll) that this level of fault tolerance is valued.
Or put another way, the domains where fault tolerance is highly valued are already so stable people do not think about it, while domains that people generally interact with just take failures as a fact of life.
You're wrong, and the poster to whom you're replying was more accurate than you.
Dave Cutler was the lead architect for Windows NT, after being headhunted, along with his team leads, from DEC to MS.
He did not work on OS/2 as you claim. He was given the OS/2 v3 project, which is to say Portable OS/2. (IBM kept OS/2 v2, which was the 386 version and which got released under that name and then later had its version number incremented. OS/2 2 was 386-only and IBM's efforts to produce a version for the POWER processors failed.)
OS/2 v3 was the CPU-independent version, little more than a plan, an outline and some header files when Cutler got it. He finished it and made it work, building it on the Intel i960 CPU, codenamed the N10. But MS wanted to distance itself from the OS/2 name and trademark.
So, when the new kernel was running and stable enough, along with its POSIX subsystem, it got a Win32 subsystem, and MS attached the Windows name to it.
Windows on N10. N10 is pronounced "en-ten". They took the initials: NT.
Windows NT is the product of Dave Cutler's labours, and it takes many inspirations from VMS.
No, it's not the same OS. No, it's not directly compatible. But there is a strong connection there.
Windows NT -- and Windows 2000, XP, Vista, 7 and 8 -- are the hybrid offspring of VMS and OS/2, as mated together by VMs' architect Cutler.
Liam P. ~ "Intelligence is a lethal mutation." (me)
You are only middlin' right. WNT is like VMS as far as the general design of the kernel goes - but most of what makes VMS the wonder that it is is missing. No versioning file system, no DCL (DEC Command Language), no Distributed Lock Manager, no clustering (WNT clustering isn't even in the same league as VMS clustering), no logical name support, no RMS (Record Management System), no sophisticated Batch and Print environment, the list goes on and on. Without these things, WNT may schedule tasks and manage memory more or less like VMS does, but does not deliver the utility and user experience that make VMS legendary.
There is no God, and Dirac is his prophet.
Of course it is, Fortran programs ran on very small processors in the past and techniques like swapping segments could make even a small microcontroller with integrated memory powerful enough given some kind of second level storage.
But it wouldn't be useful. And it wouldn't be a supercomputer which I think is the no. 1 consumer of Fortran produced cycles nowadays.
P.S. It have been spelled Fortran since forever.
that link is to dead project that tried to write an OS to the OpenVMS API, it died three years ago and there was very little code produced
to those of us who know and love this OS. Very big. VMS is secure, stable (can run for years - yes I said years - without need of reboot), and the UI is about 99% intuitive (unlike unix/linux, windows, and others). Sometimes I think I'd un-retire for a chance to work in a good vms shop ...
H.A.L -> I.B.M
?
M.S -> N.T
Spooky.......!