“Everyone is talking about Windows clusters, Unix clusters and Linux cluster. But all we are saying, is that the 20 year-old architecture of clustered OpenVMS can teach these whippersnappers a thing or two. At OpenVMS.org there’s a report about an OpenVMS cluster which handles the major processing for the Greater Amsterdam Police and naturally is required 24/7.” Read the story at TheInquirer.
OpenVMS Clusters Give Windows, Unix Thorough Thrashing
2003-12-09 OS News 15 Comments
Still, at least they buy their stuff from the right company:)
Great uptime record! It is sad, that we are getting more and more used (and accepting it) to have downtime with computers!
MS rep’s told me that W2K3 has crazy uptimes of 6 months or more !
Now that’s a real return on the investment.
Our HPUX Unix Clusters work great as well…. at our Fortune 50 airline, our databases run 24/7 to support aircraft operations and routinely have single node/database uptimes of at least 2 years at a time.
OpenVMS? Cool! But does it run Photoshop?
Is between a software built for governments/big business and one built for a toy.
I use plenty of DEC’s every day and they do work.
You can patch the system without rebooting….
Very intersting reading, but the article is short on technical content. How exactly does this all work? Why don’t you have to reboot if you update the kernel for example? Or don’t you upgrade the kernel?
That was a fun article to read. I remember those rolling VMS upgrades. I still manage 4 OpenVMS systems but we no longer have a cluster. I wish OpenVMS demanded more of my time so I’d have an excuse to get away from our 4 racks of Windows Servers. Of course VMS just runs and runs and never has a problem. The only fun I ever get to have with computers at work is when I get to do something on VMS.
I compensate for the torture that is Windows by having a Mac at home it takes about 2 hours of Mac to undue 8 hours of Windows for me. I have to do that every night to convince myself that I still like computers since I don’t get to play much with VMS any more.
Sure OpenVMS does this and it does that, and it really is a great system. I used to admin a VMS lan back in the ancient days when VMS still meant DECNet.
But here’s the problem as it is today:
1) OpenVMS is hellaciously expensive
2) OpenVMS is tied to a limited number of platforms
3) OpenVMS is even source code incompatible with every other
OS that exists
4) OpenVMS is a pain to *use*, altho it’s pretty much a joy to admin
5) OpenVMS is geared almost entirely to the business sector, it’s rare to still see it in the R&D sector as it was long ago supplanted by Unix derivatives.
The first three points have limited VMS to a small niche market and now it’s pretty much only being continued in legacy environments. The cost to migrate away from VMS is pretty high.
The best thing that could happen to OpenVMS is for it to be BSD liscensed before it becomes too obscure to be more than a fleeting curiousity. The computing community as a whole could benefit tremendously from the well implemented security system in VMS… The same one that Windows NT tried to duplicate but botched horrendously.
<history note> We all know that Unix ISN’T the best OS. The reality is that the best isn’t always the most widely adopted. Unix was adopted as the server and R&D OS of choice because it is highly portable, the source code was and is available through Berkeley and now various other portals, and it wasn’t tied to a single company and platform. Unix was and is an OS by computer hobbyists for other computer hobbyists. AT&T just simply understood the business implications of what Bell Labs and UCB CS came up with. VMS was created to fill a business computing void, but in the end failed to appeal to anyone other than the most die hard DEC fans. </history note>
“MS rep’s told me that W2K3 has crazy uptimes of 6 months or more ! “
Those were the uptimes for the beta versions of Win2003 that were being used in a production environment at MSFT as web servers, during the beta test phase. Uptimes of 6 months were considered quite good for a brand-new beta version of IIS that had been extensively re-written.
The final release version of Win2003 hasn’t even been around long enough to know what kind of uptimes it is capable of yet, so save your breath for at least a couple of more years.
For several years now, my desktop iMac, runing MacOSX at the university has been mostly a server for my course materials. It is up 24/24, 7/7 366/365.
The only times it goes down is when there is a power outage (a few hours, maybe once a year), or when I need to reboot after a system upgrade (A few minutes downtime), or when there is a major system upgrade. Even when I was running MacOS9 and before there were occasional crashes, but that only resulted in a short downtime (say, minutes), every so often. Obviously, for this type of application this is more than acceptable.
In a previous life I was a VMS manager, just before it was renamed OpenVMS. Maintenance was very expensive. Software contracts were expensive, updates were so frequent that one had to start preparing for a new update as soon as the old one was installed. IT personnel were lackeys of the DEC software machine. Most of the time a stand-alone boot was necessary, taking out the system for several hours.
At the time, clusters were already emerging, but were still in their infancy, offering protection against single point of failure issues, as well as optimizing performance, but no real assurance of continuous service. Apparently, a lot has happened since then.
The claims of continuous uptime in the article are truly impressive, but we don’t live in a mainframe world any more. The importance of clustering has decreased, not from a performance point of view (on the contrary), but from a reliability point of view. Even though mainframes can now be distributed, many other components can offer uninterrupted service, provided they give a certain amount of redundancy.
The importance of clustering has not decreased.
And it’s quite the opposite – reliability is much more important than performance. In the world of money handling software it’s paramount to perform the transaction and not how blastingly fast you did it.
2 “Dung Beetle” (Uptime, schmuptime):
You need not reboot the system for patches.
Uptime means alot if you handle critical data (usually money) or require unterruptable service (especially telecom operators).
Yeah well Tandem is still king of the hi uptime. Truth is with enough money and resources anyone can build a high uptime system with UNIX or even windows at least 5 nines.