While the mainframe is considered to be obsolete, people use PCs as servers, and the server is not obsolete. For IBM to leverage their System/360 legacy into a more reliable form of server is a reasonable enough thing to do.

And since they have chosen to distinguish emulations using just-in-time compilation from hardware emulations, not licensing software to the former, there is a legitimate competition issue to be investigated... incredible as it may seem.

Yes, the HARDWARE may be cheaper, in terms of raw computing power per dollar, however, that is NOT what makes Mainframes using OS390 a good deal, its the way the CPU's in a mainframe are licensed...

We're moving MASSIVE numbers of systems to zVM linux virtualization on our big iron and we estimate we've saved over $10M this year doing so vs licensing WebSphere, Oracle, DB2, etc on other platforms.

Also, common hardware simply can't compete when it comes to true mass performance MIPS calculations. Handling billions of transactions per quarter is not something a standalone cluster, or even grid, of x64 can do at any reasonable price. Again, hardware can be had, but licensing all those individual cores is a killer, and handling IOPS load spread across dozens of chassis (or racks of chassis) in complex grid systems requiring custom code, custom support, and custom hardware, is simply not ideal when compared to an extensible highly available big iron chassis.

The mainframe is not going away anytime soon for one other reason: legacy code. We alone have 7+ million lines of cobol code, the conversion of which is a unimaginable effort, and upon which over 2,000 other servers depend (which also would require significant rewrites of more than 450 other applications (another several million lines of editing). The development cost of replacing our code on the 8 mainframes with common hardware would be in the $100 million of dollar range, could take a decade to complete, and would require the deployment of a complete parallel infrastructure (another 150-200 million) until the new code was stable and the applicarions could seamlessly take over for the old hat. Our annual total IT budget is barely $50 million...

Slowly we're migrating services to Java and other platform independent code bases for endpoint applicartions. We're migrating to newer database platforms that could one day be moved off this hardware platform. We're deploying new DCM applications that are eliminating some older cobol code slowly but surely, but all the new stuff is still being virtualized on the Host platform. It;s a rolling cycle, and in 15 years, give or take industry trends, we might be able to give up on the big iron for most services, but they we'd still have to find good solutions for multisite replication, disaster recovery, and archiving of nearly a petabyte of data) that's growing 50% anually).