I took a university course that was specifically about IBM midrange machines. The instructor had one on his desk. I remember that the whole architecture was based around virtual machines. One assignment involved downloading a 5250 emulator and logging into some IBM mainframe to leave a message. I even got a t-shirt for that.
Unfortunately, the instructor quit about a month into the course. Just one day, we came into the room and some higher up faculty were there instead. I don't know for sure what happened, but his Linked-in suggested that he was hired by the government of Egypt. (Mind you, this was 2009) It must have been very important.
That is IBM i, aka OS/400, a completely different beast than s390. It's one of the most underappreciated and misunderstood examples of systems software in the industry. It's survived four distinct CPU changes of the same magnitude as Apple m68k->ppc->ppc64->ix32->ix64, except retaining full compatibility back to the original (and even further back to System/3x). Because everything above the analogue of the syscall layer is machine independent code, you get new benefits with even minor hardware ISA upgrades for existing code without rebuilding. This is all basically "free" for the user without any thought or downside. IBM was also doing ahead of time Java compilation to MI for ~15 years. IIRC this was one of the fastest platforms for JVM apps for many years.
The closest equivalent outside OS/400 would be like using LLVM IR for the entire system. But the memory model is also very elegant, the Single Level Store allows for storage tiering without the kind of block layer hiding that most solutions use.. closest equivalent is ZFS with a L2ARC in the POSIX world.
The lessons learned from the machine dependent part of IBM i were used to build the pHyp hypervisor in POWER4+. This is a firmware level microkernel and very interesting in its own right.
This isn't to say you should go out and port everything to IBM i, just that it should be studied by any serious OS developer and academic. While there is a good book on the design by Frank Soltis, the chief architect, it does not go into anywhere near the depth that UNIX design books did even when covering the proprietary versions. I wish I could learn a lot more about it.
It's such a damn shame that IBM makes all of this stuff utterly inaccessible to the developer, hobbyist, and academic crowds.
In the past few years I've taken an interest in big-iron systems, and I've recently started spending time with OpenVMS in preparation for release of the x64 port. I spent a lot of time digging into System i, and z/OS, but my efforts were constantly frustrated by the inability to get my hands on real software and/or hardware, or even paid time on a machine.
Shoutout for the OpenVMS hobbyist program. Anyone interested in obscure commercial OS's with serious pedigree should give it a look. HPE's OpenVMS documentation portal is well laid out and straightforward - all the information is there.
The s390 is not a midrange platform: it evolved into what we call zSeries these days and is pretty much the highest end of IBM's portfolio.
5250s were terminals used with the System/34 family, which was succeeded by the AS/400 series which is now called iSeries and moved to POWER many years ago.
Would you expect to though? I would think most of them are squirreled away deep inside the private networks of banks and such. Not something you're going to be encountering unless you work in their specific niche.
I know someone working at one of the bigger banks and they do development on some z/Arch machines. From what he's said they live in what amounts to a vault and spend their time quietly processing transactions on a secure network.
I think that's just what they tell system admins when it's time to put a machine down. "Oh yes, it's in a very quiet, peaceful place where it can process things all day long securely."
IBM puts a lot of resources into it. It helps move mainframes and CIOs eat it up.
I've heard of a lot of poc activity. When I tried it out many years ago, it really sucked. I think the use case that makes sense is when you have well defined workloads that have high availability requirements and license models that fit.
IBM dumping tons of cash into supporting it is why. Now that IBM has basically abandoned any PPC that isn't >= POWER8 and freescale/nxp/qualcomm focusing on ARM the few PPC users out there are on their own.
IBM is very good at charging astronomical quantities for legacy systems (though it seems those systems, both in hardware and in software have been receiving updates - the processors in these systems look impressive)
Basically, why rewrite an older system for (1Mi/10Mi) when you can just throw money and get a brand new mainframe?
In addition to the comment above, about sending old mainframes to deep blue pastures (so to speak), I love how your comment adds to the mystique: "I suspect ... someone is using [the platform]".
"Yes, my friend. Deep in the Dark Web, a hundred and one LANs beyond the routable Internet, there roams magnificent, ancient and magical creatures of legend - the great Zee. Quietly they tally numbers, and mete out just rewards to the poor souls trapped in this prosaic concrete reality, unable to so much as cross from our world to the Internet - let alone travel to these fabled realms of the deep, deep blue."
> "Yes, my friend. Deep in the Dark Web, a hundred and one LANs beyond the routable Internet, there roams magnificent, ancient and magical creatures of legend - the great Zee. Quietly they tally numbers, and mete out just rewards to the poor souls trapped in this prosaic concrete reality, unable to so much as cross from our world to the Internet - let alone travel to these fabled realms of the deep, deep blue."
I think I'll tweak this a bit and put it on a shirt.
When I was a sysadmin intern at my college in New York, I worked on these systems daily. We also had a very close partnership with IBM since we're located in the same city as HQ.
[1] https://en.wikipedia.org/wiki/Linux_on_z_Systems