Quote Originally Posted by kebabbert View Post
I did not know that. Do you have links on this so I can learn more?

I mean, when IBM had the TPC-C record, the Unix server costed 33.5 million USD list price. With discounts, it went down to 16.5 million USD.

But still, 16.5 million USD for ONE Unix server is lots of money. But IBM's large mainframes are much more expensive than their Unix servers. Every calculation you do on Mainframes, it will cost you. They are brutally expensive. Mainframes are something like 0.001% of all servers shipped, but account for 10% of all money used for servers. In other words, a tiny server takes a huge piece of the cake. They are expensive, yes. And have sloooow cpu performance, any modern x86 is 5x faster for a fraction of the price.

And now you say they are actually cheap? Do you have more links on this?

No, that is not the primarily advantage of Solaris. Solaris is stable, scales far better than Linux and is cheaper than RedHat.
They are brutally expensive because mainframes have *not* decreased in price as rapidly as garden-variety x86 (Moore's Law doesn't apply to mainframes). Mainframes are also not really under that much price pressure, either - mainframes are all about Very Large Transactional Volumes. Speed is not the issue with a mainframe - it's about Very Large Bites Of Bytes. It's a passenger train vs. a freight train: the mainframe is the freight train still. You *can* use a passenger train to haul freight; however, as the volume of freight goes up, it becomes less and less efficient to do so.

Take the typical transactional volume of IBM's largest mainframe - what would be required in terms of an x86 cluster to merely keep up with that amount of absolute transactional volume (not surpass it, but just to keep up)? Yes; mainframes are slow. So are freight trains. But would you use Acela Express service to move elephants to Newark? Or donkeys to New York? Or would you use a freight train (Norfolk Southern or CSX) instead?