• Please review our updated Terms and Rules here

Mainframe Myths.

The author of that page is a clown, especially when he comments on reliability. I'm going to comment on the IBM mainframes, which is pretty much the definition of a mainframe.

MVS (one of the OSes) is extremely well tested - far better than any other software in existence. It has also been around longer than most software companies.

Mainframe hardware is extremely well engineered. I/O cages are easy to work with. Peripherals can be moved and upgraded without shutting down the power to the entire system. Precautions such as error checking memory, error counters in hardware, and various fault redundant pieces are built in.

No piece of PC or mini-computer hardware will come close to mainframe reliability. If it did, it would cost as much to make.
Re Mainframe reliability

Re Mainframe reliability

I am downright proud to admit that Iave never worked on a large mainrame unless you include the XDS- Sigma 7, which I think qualifies. But I have worked on a whole lot of minis, inclding a couple that date back to vacume tube days. To say that they coould only be as reliable as a mainframe if they used the dsame quality parts, and then they would cost too much, is just WRONG! Look inside an IBM 1401. The p[c boards are brown phonellic, the components are germanium and the connectors are shocking. Then look at their maintenance logs. They typically got 4 to 8 hours out of every 24, and the big 369's and their predecesrs got the same treatment. Now look inside an SDS 920 or similar compter, o one made b DEC or CCC or HP and see how they are built. The SDS machines used silicone transistors and they ran hot hot hot, but they ran and ran and ran. We had one that ran 20,800 hours before it failed and it took just about a half hour to fix. That is MTBF/MTTR!!
At Brokhaven National Labs, where we had (among many others) two CDC 6600 computers, back to back and they shared a million 60 bit words of core memory besides their own. My boss made a casual statement at lunch one day. He siad tjhat he had studied the military's statistics on semiconductor reliability, and based on the number of silicone transistors in a 6600 (about a million) that had to all run full blast all the time top get the speed they wanted, one transistor should fail every 6 minutes. Obviously, a demonsration of the science of statistics, but on the other hand there were at least 6 CDC service people on the site at all times, and they wer very good. They received a full year of training before they ever got into the field.
The whole argument goes back to the way a machine is designed , the way it is used, the way it is maintained and the quality of the guts of the system.

Please forgive the typo's. My wife says that if I don't come down to dinner right now she will throw it out and me right after it.