Data centre managers and IT operations staff in large enterprises are now faced with a bewildering array of demands from their users, who take for granted that systems will work without any form of service interruption, planned or otherwise. At the same time the business expectations are for new services to be delivered rapidly …
While I agree with the thrust of the article - indeed, it is fairly obvious and uncontroversial stuff - I do wish people would stop writing things like
"What options are available to them going forwards?"
Probably more and better options than would be available were they to go, say, backwards, sideways or downwards. Conceivably this ludicrous piece of vacuous pomposity means "in future", but if so why not say so? And in any case the options for changing the past are dim, while the present is impossibly elusive. "Let's do it NOW! Oh wait, we can't because now has become then..."
If we are to talk usefully about technical problems, one of the first prerequisites is to use plain English, as simple and direct as possible.
Cost & app availability
While the mainframe might be super managable, there are two main problems.
Cost. Mainframes are ludicrously expensive compared to commodity x86 platforms, to the extent that an IFL (CPU core, essentially) on a mainframe to run Linux will cost you the same as a 4-socket x86 server. You can further reduce your exposure to the reliability shortcomings in x86 by clustering, whether using some clusterware on physical systems or virtualisation level clustering (e.g. KVM, ESX) so you'll get similar reliability for much less cost. IFLs for running zOS are even more expensive.
App availability. Vendors don't really sell much z-based apps, preferring Windows or Linux on x86_64 in the main, with a few still favouring Unix on RISC. What's the point in having a super-duper enterprise platform with limited software choice? While IBM will happily spout that they'll help app vendors port apps to zLinux or whatever, in reality it doesn't happen that much and you're always going to end up in a niche support corner, finding the bugs that no-one else does because everyone else is running on x86_64.
While the mainframe has its advantages and certainly has its place in some markets (e.g. most banks still run their core platform on mainframes), the shortcomings and costs need to be understood.
"...Mainframes are ludicrously expensive compared to commodity x86 platforms, to the extent that an IFL (CPU core, essentially) on a mainframe to run Linux will cost you the same as a 4-socket x86 server...."
Ridiculous. Most probably, the 4-socket x86 server has more computing power than the entire IBM Mainframe so why are you trying to move work loads to Mainframes? You should move Workloads off Mainframe, to x86 servers - you can do that by using the software emulator "TurboHercules" which gives you up to 3.200 Mainframe MIPS on a heavy x86 server. Just talk with TurboHercules, or mail them.
At my place of work we find that z/Linux only stands up cost-wise becasue we already have a mainframe, however it allows us to move big (expensive) UNIX onto mainfrane taken up by z/OS for DB/2 rather than Oracle and z/Linux. Smaller UNIX workloads are being moved off to x86_64 VMs of Win and Lin.
The big RISC UNIX server vendors are becoming very expensive bang-for-buck and are in danger of seriously pricing themselfs out of the market.