Preview: IBM z/OS v2.5 [pdf] | Hacker News Reader

Preview: IBM z/OS v2.5 [pdf]

ibm.com

I'm always curious about other evolutionary offshoots than the Unix monoculture I mostly inhabit, but this intro...

> Adaptive business and operating models, driven by accelerated disruptions, are shaping the future of enterprises today. Enterprises are embracing the next normal with an accelerated and strategic focus on application modernization, cloud-native processes, and artificial intelligence (AI), all in an effort to ensure timely and resilient business use cases and enhanced business applications for a continuous and positive user experience.

Are you still not allowed to operate a mainframe without wearing a tie?

That’s just IBM sales for you. :)

Perhaps they trained Watson to write business copy for them /s

Given the industries willing to pay for them, yes.

> operate a mainframe without wearing a tie

It's actually a safety issue. Bowties are preferred because they don't get caught in the tape reels, teletype terminals, and punchcard or paper tape handlers. Only half kidding.

It's true. When I worked at IBM long ago, the guys (Systems Engineers) who worked with the mainframes (on the "raised floor") told me that they wore only clip-on ties, so that if it did get caught it would just detach.

> The z15 provides a hardware accelerated approach to sorting using a new CPU coprocessor that can be exploited by software using the new SORTL instruction. By providing one sort accelerator per core, frequently used functions can be accelerated to help speed up sorting, shorten batch windows, and improve select database functions such as reorganization.

Having a dedicated coprocessor for sorting is extremely business.

In case anybody wants a taste of mainframe, here's the manual for the DFSORT utility: https://www-01.ibm.com/servers/resourcelink/svc00100.nsf/pag...

[Holds down PgDn for ~20 continuous seconds while not blinking]

"Reflective" and "surreal" come to mind...

More detail on the SORTL instruction: https://www.share.org/blog/peeking-under-the-hood-of-sort-ac...

Sort in memory constrained environments is interesting from a historical perspective. Mainframes did have small memory compared to data at one time. The solution was “disk sort” a process to sort files in chunks into sub files and merge the sub files into the final sorted file. Sorting is a pretty big deal in batch systems even today.

I'm annoyed by how IBM is trying to add z/OS support to open source projects. It makes sense for them of course. But for the projects there's little upside and it forces them to spend reviewer time on a system basically nobody uses.

Who wants EBCDIC in 2021 [1]? Or workarounds for yet another proprietary compiler everywhere [2]?

Open-source projects might want to think about rejecting patches for working around bugs in closed-source compilers, or patches that port to closed-source OSs that don't have a significant user base.

1: https://lists.llvm.org/pipermail/llvm-dev/2020-June/142300.h... 2: https://github.com/fmtlib/fmt/pull/2190

>But for the projects there's little upside and it forces them to spend reviewer time on a system basically nobody uses.

If someone waved a magical wand and disappeared every mainframe on the planet right now, every single financial system on the planet would collapse with it.

Why would an open source project reject commits for mainframe support? Because you don't personally use one in your business? By that logic we should dump 99% of the drivers in Linux because I personally don't have a need for them.

(yes these folks clearly have an interest in mainframe but they provide citations): https://www.precisely.com/blog/mainframe/9-mainframe-statist...

Why should open-source projects shoulder the cost of "every single financial system"? These are run by for-profit companies. They can pay for running their software, or the can open source their stuff. Right now, they try to externalize their cost and keep the profits. That makes sense for them if they can pull it off, but it's a clear downside to the affected projects.

> Why would an open source project reject commits for mainframe support? Because you don't personally use one in your business? By that logic we should dump 99% of the drivers in Linux because I personally don't have a need for them.

I think it’s less a hard rejection than a realistic analysis of the support costs. Platform support is expensive and the extremely wealthy companies using mainframes tend not to contribute resources on an ongoing basis. A really big one is CI & other infrastructure: if you actually want to have a project support your platform, the developers need to be able to run tests regularly including things like proprietary licensed compilers. Getting an open source project to support a platform really requires stepping up here since otherwise you’re just getting the untested illusion of support.

> z/OS Container Extensions provides the capability to run Linux on IBM Z softwaredirectly in z/OS. This capability enables Linux on Z application code to run on z/OSunmodified.

Instead of POSIX, we now just ship a whole OS instead as a kind of "modern" VM, given the trend to run Linux on VM across Windows, Solaris, IBM and Unisys mainframes, FreeBSD, macOS, SmartOS some of them are even certified UNIX platforms.

> Instead of POSIX, we now just ship a whole OS instead

Well, we have been saying "composition over inheritance" for quite a while... :)

Turns out, an OS that has-a POSIX is often nicer than one that is-a POSIX.

But then some OS can both have and be POSIX. It’s POSIX all the way down!

Well, the last SuS/POSIX came out in 2008. It's 2021.

It should come as no surprise that since Linux has become the defacto POSIX standard that no other OS could keep up (or would want to really).

The last POSIX came out in 2017, and it's a constantly evolving standard. It's also widely supported and Linux isn't even the most standards compliant.

Thank you for the info. I was completely in the dark about that. I thought 2008.3 was the last.

As for Linux' standards compliance, I am aware.

That’s kinda interesting in a WSL sort of way.

Last time I played with System Z machines you’d run z/Linux either direct on an LPAR or under z/VM (IIRC).

Given just how partitionable these machines are, I’m honestly not sure why you’d want to run Linux applications directly on z/OS...

I suppose if the pressure to minimize cost is really high, it could be an attractive option for shops that want to take a stab at incrementally replacing COBOL with something that runs on Linux.

If one wants to try out MVS (near indentical to z/OS) on your own pc check that out:

http://wotho.ethz.ch/tk4-/

And some really good tutorials and code from moshix:

https://www.youtube.com/channel/UCR1ajTWGiUtiAv8X-hpBY7w

https://github.com/moshix

Is it technically possible to run z/OS, or some interesting approximation thereof, on a "normal" x86 box?

Configurations that would be unusable in any sort of production scenario would probably remain interesting.

Look at Hercules, yes it is very much possible (in emulation). Performance is even decent-ish for “toy” use although of course you lose all production advantages of the hardware.

>Is it technically possible to run z/OS

Yes but not legal checkout MVS Tur(n)key:

http://wotho.ethz.ch/tk4-/

And run it with the new Hercules emulator:

https://github.com/SDL-Hercules-390/hyperion

MVS is near identical to z/OS and the Turnkey Distro has much more compilers and tools..and games already installed.

MVS is near identical to z/OS as of the mid-1970s you mean.

For a developer and a user it's nearly the same, for mainframe programming it's 99% the same, but yes no LPAR's and no z/Linux...sure. But to learn how the core of mainframes work (as of today), that's the tool/distro.

BTW LOTS of MVS code still lives in z/os

have a look at hercules and look on pirate bay for zos 1.10 adcd, ill say no more :eyes

I’ve seen many mainframes-aren’t-dead-yet news articles, and that’s clearly true—many large companies began using computers when only mainframes could do the job, have substantial investments in mainframe software, and continue to buy newer and larger mainframes to run it—but are there any new customers?

In other words, in the last decade, has IBM sold a mainframe to any organization that didn’t already have one? Do large companies founded more recently, such as Amazon, PayPal, or Tesla, use them at all?

>Authorized code scanner z/OS V2.5 is planned to provide, as an optional priced feature, an authorized code scanner of Program Call (PC) and Supervisor Call (SVC) routines for development and test environments. This scanner is designed to prevent unauthorized callers from being incorrectly granted an authorized state by detecting potential vulnerabilities in these routines with diagnostic information for remediation, as needed.

So, is this a virus scanner for mainframes?

More a program verifier.

Honestly, I wish IBM all the best. This is such a closed technology that there is no point in aporoaching it for me.

Reply on news.ycombinator.com