I am kind of annoyed that I am just now learning about how mainframes are designed in 2022, at a time when it is IMPOSSIBLE to dip a toe into the water without already having a job doing the work.
No free developer OS distributions or licenses, (so no way to run them on an emulator), no freely available emulators for modern hardware & OS releases, nothing.
Mainframe vendors, I feel, are kind of shooting themselves in the foot with this "screw the developers, they gotta PAY" mentality around this insanely expensive mainframe kit. How am I ever to learn this stuff well enough to advocate for it at work, if I can't use it and learn it in my own time? This feels like step 0 in trying to bring users to a platform, dying or not.
This article and others all show that a lot of the capabilities and redundancy requirements that we are trying to develop in PC architectures (amd64, arm64 & RISCV; are there more?) have been present in mainframes for decades, and that's an exciting thing to learn, not just because I want to write software that runs on reliable hardware, but because as a nerd this is... like ...TURBO interesting!
IBM and HP just block any newbies from doing anything of value because they're stuck in 1960s & 1970s pricing paradigms back when mainframes were all there was, so you HAD to pay if you wanted to compute at scale.
This whole situation bothers me. This is very neat technology and it's behind a massive paywall that no end-user could ever afford without a very large windfall.
As somebody developing on z/OS, I feel for you, and I am not alone. It sucks not to be able to run those systems as a hobby. (Although I don't think the pricing model is the culprit necessarily, companies are willing to pay a lot for the clouds too.)
The mainframes are also becoming more opaque as time goes on, IBM is less willing to publicly document OS and HW internals, and they are less customizable than they used to be. But that's the business trend, and life, I guess.
But yeah, the article reminded me of the new cache architecture on z16, where many processors can share their L2 caches to have a single coherent cache.
> Although I don't think the pricing model is the culprit necessarily, companies are willing to pay a lot for the clouds too.
yes, but that's a lot of small, cheap bits all put together. the minimum required to get going on the cloud is $0. with a mainframe, there is a minimum amount of hardware required and that amount of hardware costs 6-7 figures, USD$.
> new cache architecture on z16, where many processors can share their L2 caches to have a single coherent cache.
as a PC person my entire career, that is a mind-blowing thing to read.
the interconnections required to make that perform well enough to be useful must be insane. I WANNA PLAY, DAMMIT lol
Understood. I installed that old version a couple years ago. Unfortunately, I had no idea how to use it and the documentation I found wasn't the best, so I moved on to something else.
I am speaking in the present tense, not future tense.
I don't CURRENTLY know how to install z/OS. this says exactly nothing about the future or my motivation.
I do have ADHD though, so motivation to do anything is often in short supply with the current Adderall shortage. sometimes I can hyper-focus on a thing if I am highly interested and this ability has both carried me through my career and held me back significantly. surface level knowledge is pretty easy. median expertise requires a good few days of hyperfocus. a few days (6 or less) is the most I get, so expert level knowledge is perpetually out of reach, even when I am very interested in the topic :(
so I guess you've got a point. I am somewhat of a slave to this condition, and this is why it is an ADA-recognized disability.
Look for z/OS ADCD on torrent sites, older (no longer supported) z/OS versions are sometimes out there. The DASD images can be easily configured and IPLed in Hercules.
More modern version (the one that come with zPDT emulator package) cannot be pirated anymore.
I wish IBM at least made the no longer supported z/OS versions legal. No sane company would dare to run their mainframe stuff without IBM support, anyway. I think they are really shooting themselves to the foot, but hey, no executive today thinks long-term.
> More modern version (the one that come with zPDT emulator package) cannot be pirated anymore.
I don't think it can't be pirated. IBM has made it a bit harder (DRM encryption of the installation media), but nothing a sufficiently motivated and skilled person couldn't bypass. However, I guess that extra difficulty has been enough to dissuade people from actually doing it, thus far at least.
I dearly wish IBM would have a low cost box that allowed programmers to get familiar with the machines. At this point, their whole business is really reliant on existing companies. It doesn't even have to have all of the fancy hardware, just enough to get an actual environment up. This goes for the i Series as well as the z.
Not to a hobbyist, unfortunately, especially when Hercules does the job just as well (AFAIK, at least) and is free.
Compared to what you pay for the hardware and a support contract and the OS license, it is practically free.
But I don't understand why IBM doesn't just run z/VM on a mainframe and allows hobbyists to run their own VM with an operating system of their choice. Doesn't even have to be free! Five or ten bucks a month for one or two vCPUs and a GiB of RAM would be acceptable to me.
I get that due to the high profit margins[0], mainframes are a real cash cow for IBM, but it's sad they complete ignore the low-end market to preserve those margins.
[0] Last I heard, the mainframe divisions generates 10% of IBM's revenue, but 25% of the profits, so those margins must be juicy.
I agree. However, I believe that there are some within IBM and also at customer mainframe shops who like it this way because they believe there's an additional layer of security through obscurity.
> However, I believe that there are some within IBM and also at customer mainframe shops who like it this way because they believe there's an additional layer of security through obscurity.
This mindset is changing in the recent years. There is more information how to hack z/OS than ever, and IBM is trying to fix the weaknesses (for example, discourage custom authorized code, scan existing customer and vendor applications for vulnerabilities, etc.). So they are realizing that obscurity is too little protection for a determined attacker.
>No free developer OS distributions or licenses, (so no way to run them on an emulator), no freely available emulators for modern hardware & OS releases, nothing.
> It's a MVS TK4- distribution and probably 99% of a z/OS functionality plus it's Public Domain.
MVS TK4- is very old compared to z/OS. It's like comparing Berkeley Unix to Linux. Yeah, the system calls are similar but there is so much missing. Not to mention, to understand mainframes, you really need access to the whole SW ecosystem, which includes stuff like DB2, CICS, MQ, IMS.. these are building blocks for applications.
i'm sure you're right, but it's much closer than where I am today. I would like access to a proper z/OS machine so I can write some Go, but right now I'll take what I can get.
And thank you for this information, I didn't know this. I don't know anything about this stuff.
Power is Mainframe (the cores are power) too, with a OS/370 Microcode layer on top of it, with a layer of z/VM Microcode Code on top that (for LPARS etc), and then the installed z/OS or Z/VM and inside Z/VM a Z/OS, Z/VSE OR yes really another Z/VM.
Mainframe CPUs (https://en.wikipedia.org/wiki/IBM_Telum_(microprocessor) for instance) are physically different processors than Power. (Although it is likely that IBM shares the many internal design parts among both of these.)
36 comments
[ 3.5 ms ] story [ 72.0 ms ] threadNo free developer OS distributions or licenses, (so no way to run them on an emulator), no freely available emulators for modern hardware & OS releases, nothing.
Mainframe vendors, I feel, are kind of shooting themselves in the foot with this "screw the developers, they gotta PAY" mentality around this insanely expensive mainframe kit. How am I ever to learn this stuff well enough to advocate for it at work, if I can't use it and learn it in my own time? This feels like step 0 in trying to bring users to a platform, dying or not.
This article and others all show that a lot of the capabilities and redundancy requirements that we are trying to develop in PC architectures (amd64, arm64 & RISCV; are there more?) have been present in mainframes for decades, and that's an exciting thing to learn, not just because I want to write software that runs on reliable hardware, but because as a nerd this is... like ...TURBO interesting!
IBM and HP just block any newbies from doing anything of value because they're stuck in 1960s & 1970s pricing paradigms back when mainframes were all there was, so you HAD to pay if you wanted to compute at scale.
This whole situation bothers me. This is very neat technology and it's behind a massive paywall that no end-user could ever afford without a very large windfall.
The mainframes are also becoming more opaque as time goes on, IBM is less willing to publicly document OS and HW internals, and they are less customizable than they used to be. But that's the business trend, and life, I guess.
But yeah, the article reminded me of the new cache architecture on z16, where many processors can share their L2 caches to have a single coherent cache.
yes, but that's a lot of small, cheap bits all put together. the minimum required to get going on the cloud is $0. with a mainframe, there is a minimum amount of hardware required and that amount of hardware costs 6-7 figures, USD$.
> new cache architecture on z16, where many processors can share their L2 caches to have a single coherent cache.
as a PC person my entire career, that is a mind-blowing thing to read.
the interconnections required to make that perform well enough to be useful must be insane. I WANNA PLAY, DAMMIT lol
I think if it works out, it will eventually come to PCs as well.
if you were trying to dampen my enthusiasm for learning mainframe technology, you've failed, sir (or madam).
Here's someone's blog explaining how: https://famicoman.com/2018/06/28/emulating-a-z-os-mainframe-...
I want something official, supported, and free or cheap to acquire and free to use for (at least) development.
You definitely lack motivation, maybe search for something else then Mainframes then? ;)
I don't CURRENTLY know how to install z/OS. this says exactly nothing about the future or my motivation.
I do have ADHD though, so motivation to do anything is often in short supply with the current Adderall shortage. sometimes I can hyper-focus on a thing if I am highly interested and this ability has both carried me through my career and held me back significantly. surface level knowledge is pretty easy. median expertise requires a good few days of hyperfocus. a few days (6 or less) is the most I get, so expert level knowledge is perpetually out of reach, even when I am very interested in the topic :(
so I guess you've got a point. I am somewhat of a slave to this condition, and this is why it is an ADA-recognized disability.
More modern version (the one that come with zPDT emulator package) cannot be pirated anymore.
I wish IBM at least made the no longer supported z/OS versions legal. No sane company would dare to run their mainframe stuff without IBM support, anyway. I think they are really shooting themselves to the foot, but hey, no executive today thinks long-term.
100% agree. 101% even.
I don't think it can't be pirated. IBM has made it a bit harder (DRM encryption of the installation media), but nothing a sufficiently motivated and skilled person couldn't bypass. However, I guess that extra difficulty has been enough to dissuade people from actually doing it, thus far at least.
Compared to what you pay for the hardware and a support contract and the OS license, it is practically free.
But I don't understand why IBM doesn't just run z/VM on a mainframe and allows hobbyists to run their own VM with an operating system of their choice. Doesn't even have to be free! Five or ten bucks a month for one or two vCPUs and a GiB of RAM would be acceptable to me.
I get that due to the high profit margins[0], mainframes are a real cash cow for IBM, but it's sad they complete ignore the low-end market to preserve those margins.
[0] Last I heard, the mainframe divisions generates 10% of IBM's revenue, but 25% of the profits, so those margins must be juicy.
This mindset is changing in the recent years. There is more information how to hack z/OS than ever, and IBM is trying to fix the weaknesses (for example, discourage custom authorized code, scan existing customer and vendor applications for vulnerabilities, etc.). So they are realizing that obscurity is too little protection for a determined attacker.
Distro:
https://wotho.ethz.ch/tk4-/
Update to Distro:
http://www.prince-webdesign.nl/index.php/software/update-on-...
Emulator:
https://github.com/SDL-Hercules-390/hyperion
It's a MVS TK4- distribution and probably 99% of a z/OS functionality plus it's Public Domain.
Or VM/370 another Public Domain OS from IBM:
http://www.vm370.org/vm
Or even better you run MVS TK4- inside VM/370...for the real taste ;)
But if you really really want z/OS (i have no clue why you would do that) you click here:
https://www.ibm.com/it-infrastructure/z/education/zxplore
(Formerly know as Master the Mainframe)
And some yt's to have a little base-knowledge (mostly MVS and VM/370):
https://www.youtube.com/c/moshix
MVS TK4- is very old compared to z/OS. It's like comparing Berkeley Unix to Linux. Yeah, the system calls are similar but there is so much missing. Not to mention, to understand mainframes, you really need access to the whole SW ecosystem, which includes stuff like DB2, CICS, MQ, IMS.. these are building blocks for applications.
But you are right, if you want that, got to zexplore. If you want a taste of Mainframes then MVS and VM/370 are a excellent starting point.
BTW: If you understand "Berkeley Unix...whatever that is" then Linux is a piece of cake...not that ~much has changed.
And thank you for this information, I didn't know this. I don't know anything about this stuff.
ehh...well i let you figure it out...
https://www.youtube.com/watch?v=5jOlO1vSBzE
>we run MVS 3.8 on top of VM/370, on top of zLinux, on top of z/VM, on top of an LPAR, on a real iron IBM mainframe.