OK, some twenty years after InfoWorld editor Stewart Alsop announced the death of the mainframe (see the New York Times article here), it's time to put the poor thing out of its misery. I declare the mainframe finally DEAD. RIP. Dead parrot time.
Of course, if you want a server that is never unavailable (well, one where 5 minutes downtime a year, including "planned downtime", is worrying); one which can handle multi-tenanting with no possibility of one tenant affecting another's service; which can run several programs at the same time with no risk of something low priority stopping something higher priority working; and one that is capable of running, say, an ATM business for the whole of N America; then you might still get a zEnterprise from IBM as an alternative to, say, an Oracle Exadata box. But you didn't ask for a mainframe, you asked for a service with certain, quite testing, characteristics. Whether you end up with a zEnterprise or an Exadata box (either in-house or as a managed service) simply depends on which can meet your SLAs for your particular workload and needs; and, if both can, which costs less-over the expected usage lifetime (don't just count acquisition cost).
Actually, these days you may not even know what box you got-and why should you care? You probably bought a service, and why should you care what it runs on, as long as your SLAs are met?
So, the mainframe is truly dead at last, because no-one around management level really cares if the computer behind their service is an overstressed Apple Mac or a cool zEnterprise running Linux on a blade. As long as the SLAs are met, why should they? Still, I don't expect the Mac to replace the zEnterprise any time soon. Someone techie in the service provider still has to choose a hardware platform that makes meeting its SLAs at least plausible....
The mainframe is dead. There will be a period of grieving, but then you'll just get on with living. The mainframe is dead. Long live the reliable, scalable cloud server!
PS, there will, however, be consequences from this sad loss. It would be a good idea, in this new world, for all your management utilities to be cross platform-which means that they should be able to manage the user experience (including end-to-end performance management) across all possible platforms and through into the cloud. And the possibility of SaaS delivery might be appropriate too. A message that seems to have got through to CA Technologies (by the way, check out Nimsoft) and Compuware (with its focus on Gomez for APM and its recent Dynatrace acquisition), at least.
There's also a risk, if you're in a company that already has mainframes (and that will continue to run "enterprise cloud servers"), that your existing mainframe staff will be feeling a bit lost and insecure without the mainframe name. You don't want to lose these people, as they represent many man-years of knowledge of the business and of your technology infrastructure. You should be making them feel valued as mentors, in a skills-transfer program (perhaps using something like CA Chorus)-your distributed platform support staff can probably learn something about running very big and very reliable systems from them.