On Thu, 16 Jun 2022 at 04:46, Timothy Sipples <[email protected]> wrote: [...] > Now there's yet another great option: the IBM Virtual Dev and Test for z/OS, > generally available THIS FRIDAY. The announcement letter is available here: > https://www.ibm.com/downloads/cas/US-ENUS222-251-CA/name/US-ENUS222-251-CA.PDF > And the landing page is here: > https://www.ibm.com/products/virtual-dev-and-test-zos > > This product also provides virtualized z/OS instances on IFLs for > development, unit testing, and training. You will need at least 2 IFLs (on an > IBM zSystems or LinuxONE server), although they can be potentially shared > with other > workloads. Of course you'll need some memory and disk space. The latter can > be any reasonable Linux-attachable storage. (It does not have to be > specifically ECKD/FICON-attached.) IBM Virtual Dev and Test for z/OS licenses > are available via IBM Passport Advantage in your choice of one-time charge > license (with annual Subscription & Support) or monthly license (with monthly > Subscription & Support). You can start with monthly then switch to one-time > charge if you wish.
It would be really nice if IBM would explain what this product *is*. I appreciate that a certain amount of marketing fluff is always necessary, but even for IBM this is remarkably heavy on the fluff and light on the details. All we really know is that IBM seems to be doing what it went to court with Neon Systems over - running arbitrary z/OS work on "specialty engines" on z. Seems to me it's roughly one of two things (but doubtless there are many possibilities I haven't thought of): 1) It's zPDT compiled for a target of zArch rather than Intel i64. 2) It's some other virtualized - probably container-based way of dispatching z/OS work on IFLs, with user-friendly infrastructure to provision (dare I say "orchestrate") VMs and their content. I think (1) is the more likely based on what can be seen. We know that IBM has the JIT compiler infrastructure to target zArch, so presumably it wouldn't be a huge thing to just recompile zPDT and go. The problems most commonly heard about zPDT are with the licensing infrastructure and Linux library versions, and not with the emulation itself, which seems to be very robust. Of course no amount of JITting is going to run as fast as just dispatching on the native iron, but presumably the tradeoff of cheap and fast IFLs vs some overhead works out well. If it's (2), then this must play into the zCX and related container stuff that would otherwise seem to be plugging along in a parallel universe. But the obvious question that arises is why not use one of the several existing schemes for spinning up virtual machines on z - whether zVM and its many tools, or UNIXy tools to do the same? These are fairly mature at this point - certainly the zVM-based ones. I think (2) is far more the Future than running zPDT on zArch. Maybe it's just the traditional IBM Way of having competing teams with a periodic Winner and some Losers, and this battle isn't over yet. One more little confusing item I noticed: In the announcement letter under Software requirements Operating systems we see: • Ubuntu Server 22.04 LTS for IBM Z and LinuxONE; IBM z/Architecture(R) • Ubuntu 16.04 LTS, 18.04 LTS, and 20.04 LTS; x86-64 hardware • Red Hat(R) Enterprise Linux (RHEL) Server 7 and 8; x86-64 hardware This is just weird. What are those two Intel Linuxes doing there? Surely, surely this doesn't in any way involve i64 emulation on real zArch hardware? Tony H. ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
