That's what I figured. I have to wonder what is the attraction to that vertical. I guess the upside of a mainframe is that they are vendor-managed. One thing you can say about most banks below the investment bank level is they are fucking terrible at IT.
The main benefit is raw throughput latency. Mainframes tend to have crazy fast IO busses and multiple redundancy in the same cabinet. Plus you can run just about anything on them now. Used to be you'd just create batch jobs with COBOL in zOS, but now you can run Linux, AIX, VMs, and they even have dedicated Java CPUs.
Thanks. That's a really interesting insight. So while pretty much every other workload is moving to a cloud or even to a cluster scheduler on the cloud with metric fucktons of software abstraction layers between the work and the hardware and all IPC commo being mostly over IP Linux VMs on a mainframe would likely kick the shit out of a cloud for raw performance due to "local" IO.
(post is archived)