roscoe5 asked: >how do you see the future for mainframes?
>Increasing, steady, declining, . [Editorializing ahead!] As usual, "It depends". There are fewer mainframe shops than there were, but more usage. A simple example: consider payment processors, many (not all) of whom have at least some IBM zSystems. Recent consolidation there (multi-billion-dollar deals): FIS bought Worldpay; Fiserv bought First Data; Global Payments bought Heartland and TSYS. Seven companies are now three. So there's your "fewer customers". Meanwhile, of course, transaction volumes at these and other committed companies are still growing. So there's your "more usage". What I assume will happen over time is that along with continued consolidation, some shops will move off because some bright young spark is convinced it will be better. Doesn't mean they're right, but that doesn't mean it won't happen. Even when that doesn't happen, various other evolution will chip away at some usage when it becomes (or, again, SEEMS to become) wiser to move some processing off. And essentially nobody is moving new processing TO the mainframe, because reasons. At the same time, while the smarter companies get it and see no reason to move, there aren't a significant number of new mainframe shops. A few LinuxONEs, but zero new z/OS or z/VM or z/VSE or z/TPF shops. I don't see any way this trend will reverse, though by the same token I don't see the mainframe going away anytime soon. It will just become more and more of a niche market of large systems-sort of strange, "big niche"! The aging of the mainframe community isn't help, of course. In some ways it looks (at the right distance) like a return to the early days of computing, where big iron shops were few but serious. Of course then it was big iron vs. nothing; now it's big iron vs. racks, cloud, etc. Maybe AI will reverse this trend, though I personally don't see it. Like many current Internet services, AI usage looks like it will largely comprise two things: building the LLM (which is expensive but not real-time and can retry on failure, so might as well use cheap MIPS) and then querying/using the LLM (which is real-time but not critical, so might as well use cheap MIPS). Plus there's been a whole shift from "computers should work" to "computers should *mostly* work". When your Google search fails or your Instagram page doesn't load, you shrug and try again. When your credit card transaction doesn't go through, neither you nor the merchant just shrug. But you're starting to-when it's a website and the transaction fails, you scowl but try it again, and if it works that time, you forget about it. We're being conditioned to accept mediocrity, and AI in its current incarnation doesn't appear to be ready to reverse that. It's depressing. Meanwhile, a colleague happened to send me this: https://www.itpro.com/cloud/cloud-computing/cloud-computing-or-mainframe-why-the-pendulum-might-be-swinging-back-in-the-age-of-hybrid-strategies-and-generative-ai which is a bit more cheerful, albeit more "the leak is slowing" than "things are improving". ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
