The evaluation release of HP OpenVMS Version 8.1 for Integrity servers based on the Intel Itanium processor is out. OpenVMS Version 8.1 includes native compilers, and a substantial suite of functionality that ISVs and customers are eager to use, including clustering and a wide range of development tools and integration technologies.
what might be the advantages of using OpenVMS vs. *nix?
telnet: dahmer.vistech.net
web page http://dahmer.vistech.net/
None, really, unless you enjoy bashing your head against the desk.
there are still a lot of legacy apps that could run on it, especially if they’ve never been ported and you only need to run them for the port process.
VMS worked well. But had some _terrible_ support for networking as the TCPIP stacks matured elsewhere. And DEC really knew how to market and sell stuff, yes? Sorta like Apple IMO, ahead of its time in many areas and sold off the best of the stuff.
I’ll try this one at work as soon as I can find the time!
VMS has been what IT is moving towards for almost 20yrs now.
It’s stability is A.W.E.S.O.M.E!
The VMS groove master at work just smiles when we talk new features in Linux/*BSD or even windows. We don’t ask why nowadays… we know the answer – “VMS could do that in the late 70’s – maaaan” in the worst hippie surfer style imaginable.
Now, WHY don’t I have a couple of “spare” Itanium servers at home?? It’s winter damit! I need the extra heat!
Name one thing VMS does that Windows XP doesn’t…
Let’s us see:
* File versioning. Until you’ve used it you won’t appreciate how useful this feature is. It’s my most missed feature of all time.
* Native indexed file system. You can write very useful record-based systems in a heartbeat.
* Completely cpu-type accounted processing (Event-based priority processing). VMS has amazing performance per clock cycle because if the system is completely queued on blocked calls, the system literally does nothing except for the clock update.
* Clustering that works, including galaxies that make clustering look like a toy.
* DCL is still an all time favorite shell. Name another system that can mix and edit heterogenous, indexed data files on the command line.
You really just have to see the internals of how the system works to be impressed with its capabilities. VMS is based largely around the needs of real data processing. If you’re wanting to see flashy start buttons and little pieces of clip art dancing on the screen, VMS is not for you. But, for serious data processing, the system screams and scales massively. I miss working in the system since I left my old company. Stuff that I used be able to do on the command line I have to use or write an application. Though, I’m curious with the conversion to Itanium what features may have gotten dropped. Some of the systems relied on specific abilities of the processors and buses.
Bel, the mostly sane..
i belive the uptime record for vms is 18years without a reboot
i would like to se other systems do that.
ofcourse for the newer ones we only haveto wait about 17years
to se if they pull it off
att: Beldraen
I guess VMS has many interesting features. Unfortunatly I have never tried it (I would like to, but don’t have money/equipment/time).
Can you point me to any places where I can learn more about how the file-versioning are implemented?
What is “native indexed file system”, I don’t understand that ?
I am working on a ruby-centric operating system.
http://ros.rubyforge.org/
—
Simon Strandgaard
I have been a VMS/Unix System Administrator for the last 17 years and all I can say is “wow”.
Working on both sides of the spectrum gives you a great appreciation for what each can do.
From my admin and C developers point of view, VMS outperforms Unix in reliability/security/system management, accounting, problem detection/resolution, etc.
On the bad side, The older ODS-2 file system disk and directory structure is different under VMS though and that tends to put off Unix/Windows users.
Okay, It freaks them out.
That being said, there are somethings that the Unix variants do better.
Unix shell command line editing such as sed, ed, awk, grep, etc are great.
What made VMS unattractive in the early days was it very expensive while colleges and universities got Unix essentially free.
cost an arm/leg on propriatory hardware versus free = tough choice
The only people using VMS nowadays are hospitals, military, air traffic control, financial institutions and similar places that just cant have outages.
Organizations who know the value of 99.999+ availability know about VMS and use it but it is not a OS for the home.
Or is it.
Former DEC engineer Dave Cutler wanted to make a desktop version of VMS under project emerald but it was killed.
Cutler then went to Microsoft and ended up porting alot of VMS’s features to Windows NT and its child OS’s 2K and XP.
NT’s kernel was basically revamped VMS.
So.. VMS lives on the desktop in away.
I was downtown Vegas last year, drinking at some crappy bar and I get into a conversation with a with the guy next to me. Apparently, he was one of the original OpenVMS developers and he talked about how it killed him that this great operating system had been pretty much ignored by Compaq/HP. He told me that he pleaded with them to release it as Free Software – just to breathe life back into the operating system.
is this damn box still runing?
Last time I heared from it was more than 4 years ago with an uptime of 12 years. If I remember correnct it’s located in Irland.
On a side, but very significant note, VMS is where clustering got invented and implemented.
The various Unix variants try to emulate some of the power, flexibility, and fault tolerance that VMS clusters have but never really pulled it off.
An interesting exception to this is TRU64 Unix.
Small wonder since DEC and then Compaq made both.
What is sad is TRU64 is going into the mist now that HP has taken over.
HP is picking the best parts of TRU64 (including clustering)
and sticking them in next generation versions of HP-UX.
If you’re looking for implementation specific information, there was a VMS internals book at my workplace. I didn’t think get a copy or get the name of it. Look for things done by Bruce Ellis. He’s a VMS god. As for how it worked? Pretty simple, unlike traditional files that are <filename>.<ext>, VMS adds on ;<version>. So, you may see a files like this in a DIR listing:
foobar.com;3
foobar.com;2
temp.data;5
temp.data;2
temp.data;1
Opening and saving automatically creates a new version. If I realized that I screwed up foobar.com;3 royally, then I can just delete it and go back to version 2. If I decide that I want to clean up my directory from old versions, I just say PURGE /KEEP=5, and I’ll keep the five most recent versions of all files.
As for indexed files, it means that a key (actually up to eight) may be specified in the file, as well as record size and other attributes. When programming, instead of having to use ADODB (or Microsoft’s flavor of the month in software APIs and underware), I could open a file and read for a record using a key in the language. In other words, the idea of a “database server” is meaningless, to a degree, in VMS because record based access is built into the file system itself. Net result? Blindingly fast access. As in for general processing we found it to be 6 to 10 times faster than Oracle and on complex programming to be 100’s of times. Granted, Oracle is meant to work with people who do not have the programming experience to develop and implement their own systems; thus, it has to be generic and that genericy costs performance. The database we ran could generate a full report in HTML from the needed records in about 1 hundreth of a second, compared to tens of seconds for Oracle.
Good luck on your project!
Bel, the mostly sane..
I think I read somewhere that NTFS has versioning – it is just not exposed. But maybe not.
I wonder if HP has any regrets porting to Itanium – now that it appears that Intel is going to dust off Yamhill to counteract AMD 64 possibly leaving Itanium to the same fate as the Alpha.
http://dahmer.vistech.net/defcon.txt
http://www.pottsoft.com/home/vms/VMS_TUTORIAL.HTML
HP’s only reason for the merger with Compaq as to get rid of Alpha as competition for IA64.
But the best is that there are still plans to further develope Alpha if IA64 doesn’t get a significant marketshare (and I realy hope this will never happen)
while uptime is NOT an important issue here, if you’re going to use VMS it’s going to be for some of the unique features or legacy apps, i am curious to know what the uptime record server has actually been doing for the past 18yrs, if it’s connected to the web or running a bucketload of heavy services then i’m impressed, otherwise, i suspect your average debian box could sit in the corner and hum to itself for 20+ years with the right power backup configuration.
not necessarily on x86 though…
No one has yet topped VMS in the cluster arena. CHeckout the following article from TheInquirer.Net:
http://69.56.255.194/?article=13002
There is a OpenVMS clone called FreeVMS
yeah, but FreeVMS is deader than BSD
and AFAIK it’s just about featureless.
Imagine that Bill Gates said, “Windows sales are projected to go down, so I recommend that customers switch to Linux” and in response, customers from operators to CEOs visit Bill Gates to explain how important Windows is and how Linux lacks essential capabilities, and so Bill says, ok, we’ll support Windows as long as customers want to buy it, and then Bill lays off half of Windows engineering. And a decade latter, operators to CEOs are still visiting Bill Gates to plead for him to just sell Windows.
In other words, customers value VMS far more than the management of DEC, Compaq, or HP over the past decade plus.
Further, the management of DEC, Compaq, and HP have preferred to sell products that generate less profit than VMS because sales of VMS were forecast to reach zero in 1998.
Oracle used VMS as the development platform until a DEC VP told Oracle to switch to SunOS. Actually, he didn’t say SunOS, he said unix, thinking that Oracle used VMS because DEC was so great and he expected that Oracle would switch to ultrix which had a fraction of the VMS market share.
In other words, VMS customers have been beaten to death so DEC, Compaq, and HP can pry VMS from their cold dead hands.
att: Beldraen
I have copied your description to our wiki page. If you know about other interesting techonologies in OpenVMS then you are welcome to add it
http://ros.rubyforge.org/wiki/wiki.pl?BestFeatures/OpenVMS
BTW: Happy new year.
—
Simon Strandgaard
VMS 7.3-1 and higher implement the ability to used an enhanced I/O cache called XFC.
A system admin can selectively allocate P2 memory space to cache frequently accessed files.
If an update to a file is required, the file is updated to disk as well as memory.
The result is blazingly fast database reads.
I have seen Oracle RDB production processes speed up by up to 40% using this caching.
Please note that the allocated P2 memory space dynamically shrings/grows depending on what memory resources the system needs.
Details on XFC can be found at
http://h71000.www7.hp.com/doc/731FINAL/6017/6017pro_081.html#displa…
If you’d like to see the VMS command line there are a few public and semi-public machines. Hobbit, Manson, etc. And a few dcl emulators although most may be commercial. I think a free one is “dcl lite.”
Interestingly enough, one of the main machines for student access at a local university was still running VMS in ~ 1996. I figured most every university was running Unix by that time.
As a *NIX admin/C hacker who has spent far too many (20+) years in way-way too close proximity to VAXen and VAX/VMS bigots, my first reaction to this news was much like my reaction to “sugar-free, caffine-free Coke” – “Why? What’s the point? Could ANYTHING POSSIBLY be more useless?”
I’ve written some serious VMS systems programs, because I “got” C, and my VAX-BASIC programmer cow-orkers didn’t; and the VAX admins only knew DCL (and maybe not all that well). The assembly-language API to VMS, with its indescribably complex “union of structs of pointers to unions of bitfields” behavior makes the MS-DOG 3.2 API a model of clarity and simplicity. To accuse VMS of being baroque, byzantine, overly (complex, wonderful, glorious) would be a major understatement. VAXMACRO is the ONLY way to go here – and you DON’T want to go there.
My favourite oxymoron has always been “OpenVMS”. Like “broccoli pizza”, this is a notion fundamentaly at war with itself. When the VAX/VMS bigots I work with explain that “FOO /NOEVERLASTINGGOBBSTOPPER /DWARVES=7” is “intuitive”, milk often squirts out my nose. When GNU makefiles refer to VMS as the “Vomit Making System”, I smirk knowingly.
So, I ask myself, what is HP(COMPAQ(DIGITAL)) up to here?
Compaq bought Digital for their field-circus group (and immediately downgraded VAX (which was already on the wane – for Alpha)), then HP buys Compaq (for no clear reason), and (being the other driving force behind IA-64) downgrades Alpha , in favour of Merced/McKinley/Itanic/Whatever.
Suddenly all the businesses who’ve been running VMS 4.2 on VAX 9000s since 1985 have no upgrade path (as if they’ve been urgently pondering this recently). So what’s the HP nee Compaq nee Digital sales farce supposed to sell them, just in case they ask?
Ta Da – “Open-VMS on Itanic^Hium”.
—————————–
Quick scorecard/glossary:
VAX processor – nice CPU, in say, 1982. Big CISC instruction set, optimized for writing complex and womderful operating syatems.
VMS operating system – nice OS, in say 1985. Hopelessly, irretrievably tied to VAX hardware. Sad attempts to make it run on ALPHA cause laughable rename to “OpenVMS”.
ALPHA processor – a turbine engine, invented by a company which made buggy whips. Way ahead of its time, doomed to failure since birth.
PA-RISC processor – the “crewman in the red shirt”. A pre-ordained victim of “friendly fire”. Truth be told, a pretty great CPU.
IA-64 processor – AKA “{[A-Z][a-z]*}ium” – the last, best hope of mankind (or something very much like that). Delivery date is “Real Soon Now”. Working versions are expected “Soon After That”. The (proposed, future) physical implementation of the phrase “world hegemony”.
—————————–
So, my original question was: “Why? What’s the point? Could ANYTHING POSSIBLY be more useless?”