[krylon]

I am not sure if this true for current hardware, but once upon a time the mainframe CPUs contained duplicates of each functional unit that ran in lock step; after every computational instruction, the results of the units are compared; if they differ, the CPU repeats the instruction once. If the results still differ, it "calls in sick" to the operating system.

The OS then brings a spare CPU online and transfers the program that was running on the failing CPU to the new CPU, takes the failing one offline, and, depending on your service contract with IBM, calls home for a replacement CPU. The program does not even notice something went wrong. The next day an IBM service technician rings your data center's doorbell and replaces the faulty CPU, all without taking the machine offline.

That kind of resiliency and redundancy runs throughout every aspect of the system's design. If you can afford it, having a mainframe be your single point of failure is not too bad.

[Rjevski]

This is definitely impressive, but most of it could be done with commodity hardware and open source software. Let's say I have a task queue with objects like (1, 2) and my task is to add them together and push the result in a new task queue - what's preventing me from doing that twice to make sure the end result is correct?

Sure, I have to implement it manually, but at least my screen will no longer covered in vomit because of all the buzzwords.

[krylon]

My point was that a mainframe could be a single point of failure, but it engineered to such a degree of reliability and availability, that this is unlikely to be much of a problem. Plus, you can cluster them, you can even build clusters spanning several data centers.

From a technology standpoint, I think they are amazing machines.

The problem, of course, is that they are ridiculously expensive, which is why "open systems" (often meaning PCs running Windows or Linux, but also proprietary Unix system) have replaced mainframes in many places. In a way, I think it is fair to say that Google did something akin to what you propose. (I am told, that proprietary software for mainframes is actually so ridiculously expensive that the already-ridiculous cost of the hardware is not that much of a concern, actually.)

As for the buzzwords, if you find a place where you are safe from them, send me post card, I might move there, too. ;-)

[snowcrshd]

Thank you, this is the kind of information I was looking for! I still would like to see some real world comparisons, though.

It sure sounds pretty impressive from your description!