Re: Tech is the biggest problem facing archiving • The Register

On Apr 9, 2013, at 3:51 PM, MG <marcogb at xs4all.nl> wrote: I have. I originally ended up looking up its predecessor (S/390) when IBM started supporting S/390 Linux back in... the late '90s? Early 2000s? Nope. No, but I don't work at a bank or a stock exchange. I once worked at a health insurance company that might have had one, but I certainly never had direct contact with it. I don't read current computer publications. Blogs, yes, I've seen it mentioned (usually in conjunction with the software patent nonsense they've been pulling with Hercules). I know there are public forums on zos.efglobe.com. It appears to be down at the moment, but I haven't seen any explanation of why. Same place: zos.efglobe.com has a public-access system (running z/OS 1.6, which is a bit outdated). It's the only free one I've seen; any others are usually rentals of time on a system because the normal assumption is that you'll be using it for commercial purposes. My two cents (and PLEASE don't make this all emotionally-charged, because it isn't): The mainframe market has established customers (mainly banks) who will continue to buy IBM mainframes because the software that they've invested DECADES into still runs on modern IBM mainframes because IBM is heavily invested in backwards compatibility. They fund IBM's mainframe budget because mainframes are a low-volume, niche industry with high margins. They're analogous to supercomputers, except they trade raw number crunching ability for sophisticated high-bandwidth I/O systems that can support incredible sustained data rates; it's a feature set that no one except very large businesses tend to need. Those very large businesses will train entry-level programmers on their mainframes, so there's very little need for a presence in universities; additionally, the sales channels are essentially established through pedigree. If you have enough money and the requirement for high data throughput, IBM will cold call you. To use your analogy, it is actually a lot like a door-to-door vacuum salesman, except in this case, the salesman already knows you need a vacuum and is probably one of the only people in the world who can sell you the right one. And to address a previous point: yes, it is the same thing that the NonStop and VMS people say when they point out that they're used around the world in stock markets and banks. They're also all mostly correct; NonStop has about the same customer type as IBM does for its mainframes. VMS is maybe a step below that, but there are still a lot of organizations using it because they have a lot invested in VMS-based software (or because it's actually the best tool for their particular task and someone who specified the system actually knew something about VMS). Admittedly, that's basically three major cases of vendor lock- in, which I'm generally not a fan of because it leads to stagnation. Mainframes aren't any more dead than Formula One racing is, but it's understandable that you might think so. The same thing COULD be done with a giant cluster of x86 machines, but IBM has spent 50+ years making a single architecture with the total package of reliability, scalability and pedigree. It's simply an easier call for gigantic enterprise operations to buy a total package that has it all than try to, say, make Linux be fault-tolerant enough to run a whole international bank. Some banks do it, but plenty of others don't, and it doesn't take much to sustain an industry that runs on that much margin. - Dave