I like Mainframes. I like and understand their architectures and behaviours (try saying that about a Windows server with a straight face – Windows invented the mindless “switch it off and on again” meme, not really an option if you are running the books of the bank). The architecture is rock-solid and well-designed, without obfuscation, which is one reason why mainframes are so resilient and robust.
Other reasons include rock-solid virtualisation – I used to work on VM/370 (operating_system), a virtualised operating system provided as an alternative to MVS, the “native” OS for IBM 370 mainframes, back in the 1980s and it was about 10 years old even then – and rock-solid parallel programming (Parallel Sysplex), with pre-emptive multitasking (which means that, basically, a rogue task can’t grab the CPU and refuse to let go and kill parallel processing). By “rock-solid” I mean designed in, not bolted on as an afterthought.
Every so often I am reminded that mainframes are still around and still processing a lot of the world’s commerce. BMC has just acquired Compuware, which complements BMC’s mainframe management capabilities with Compuware’s mainframe DevOps capabilities. I am saddened that there is one fewer mainframe vendor now, but the merger makes sense to me and there is still Broadcom (which acquired CA Technologies) to compete with. There’s money in mainframes, because there are still applications where only a mainframe will cut it.
But hold on, in 2020 the mainframe really is no more. This having been announced just about every five years since about 1990 – even I did it, ironically, in 2011). IBM has done a superb job modernising its Z mainframes, so that they are “just” another large, performant, highly resilient “super” server – which supports cloud delivery (Mainframe as a Service) really well. Companies like BMC have enabled management of the mainframe as part of the general business IT infrastructure and companies like Compuware have integrated mainframe development/maintenance into modern development practice.
And, there are people like Unisys, building things out of x86 servers (Unisys Clearpath) that have all the characteristics of a mainframe.
So, hey, The Mainframe is Dead. Long live the SuperServer!
Forget about the old “Migration off the Mainframe” messages. Somehow, they never quite delivered (I could write several long papers in why this is). The message for the 2020s is “Run the Right Workloads on the Right Platforms” and the right platform might well be the mainframe, if security, resilience and ultimate workload capacity matters. And sometimes it won’t be the right platform but abandoning the mainframe should be a fact-based choice driven by workload, not a religious decision.
Which rather implies, of course, that you do understand what workloads you are running for your business, and what their characteristics are; and that you are actually managing your SuperServer, not just taking it for granted. Even legendary SuperServer security is only achieved if you have appropriate security policies, part of a holistic security environment, and they are properly implemented. There is still work to be done.