OpenShift Can Be The New PASE For IBM i Shops
May 1, 2023 Timothy Prickett Morgan
If you can’t beat ‘em, encapsulate ‘em. That is not the official rallying cry of the IBM i platform, but that certainly has been the philosophy of the Private Address Space Environment that was cooked up by the techies in IBM’s labs in Austin, Texas, home of the RS/6000, and Rochester, Minnesota, home of the AS/400. The marketing people eventually changed its name to the Portable Application Solution Environment, which we know as PASE either way. We have a great PASE, but it is time to make a new one that is more relevant for the time. We are not in the Unix systems era, but the Linux systems era.
I’ll even give you a name for it, IBM: Portable Linux Enterprise Application Software Environment, or PLEASE for short.
In June 2000, as you can see in announcement letter 200-167, the headless AIX runtime environment known as PASE was embedded in OS/400 V4R5. That release was one of the biggies in the OS/400 and IBM i line, and demonstrated that the system architects who gave us the System/38, the System/36, and the AS/400 were still working along the vast hallways of Fortress Rochester. OS/400 V4R5 was the first release to allow for the mixing of multiple operating systems on a single hardware platform within the relatively new hypervisor that would eventually be known as PowerVM. It also allowed up to 24 partitions on a machine, double that of V4R4, which actually used a variant of OS/400 to host OS/400 partitions.
That V4R5 operating system also included Integration with Windows Server, a feature that allowed for Windows NT Server 4.0 or Windows Server 2000 to be deployed on integrated Netfinity Servers, which were previously known as the Integrated File Serving I/O Processors that actually made their debut way back in 1995 running IBM’s OS/2 LAN Server, Novell’s NetWare, and Lotus Development’s Notes.
This was another example – from a long list of examples – of IBM grabbing what was interesting and useful at the time and integrating it with its midrange platforms. The first obvious integration being the experimental merging of the System/36 on top of the System/38 in the wake of the failed “Fort Knox” converged mainframe-minicomputer effort in the 1980s that ultimately lead to the launch of the AS/400 nearly 35 years ago.
IBM doesn’t integrate for the sake of integration, but brings something into the OS/400 and IBM i stack with the explicit goal of either making a thing easier to deploy or use, or better still, absolutely invisible to programmers and system administrators as it is deployed and used. So it was with PASE, which was initially based on a headless subset of AIX 4.3.3 that allowed 32-bit AIX applications to run natively on the Power-based servers that ran OS/400 and then IBM i. (Remember this was ahead of the formal convergence of the AS/400 and RS/6000 platforms.) PASE has grown to cover 64-bit processing and is not only used to bring third party – and often open source tools – to the IBM i platform with native OS/400-style controls and APIs, but has been used by IBM itself to converge the AIX and OS/400-IBM i versions of native system programs.
The first time that I can remember that IBM itself used PASE to lighten the programming load at Austin and Rochester was when it deprecated the native OS/400 TCP/IP Ethernet networking stack and just dropped the AIX version in PASE. Did any of you notice? Did any of you care?
Nope. And that is the whole point.
So, if IBM creates a PLEASE runtime that runs containerized Linux workloads under the covers of IBM i but is really a full OpenShift Kubernetes environment that looks native to IBM i for a Linux workload, will you notice? Will you care?
Nope. And that is the whole point.
So that is what we think that IBM should do to both proliferate the OpenShift stack across its enterprise customer base and to help IBM i shops get more workloads off X86 iron and onto Power iron, thereby making the Power Systems platform more central to the business and thereby creating a virtuous cycle that improves computing at IBM i shops – and AIX shops, which also now need a PLEASE Linux container runtime of their own – and thereby drives more sales of Power Systems, and so on and so on.
I know what you are thinking. This is like the Virtual I/O Server on steroids, which is going to be an anathema to IBM i shops, who have no great love of VIOS, as we all know. VIOS. Of course, is another slice of AIX that provides a virtual I/O driver for peripherals so IBM Rochester does not have to create unique drivers for the IBM i environment. (If PASE is for compute, VIOS is for I/O.) The difference between PASE and VIOS is that one is essentially invisible and the other is definitely not; that may be by good design in terms of PASE, but perhaps there is something about I/O that means VIOS can’t be so painful and companies end up having to do Unix stuff that is basically alien to them – and they don’t like it.
Let’s talk first about why IBM needs PLEASE. In May 2019, more than 1,000 customers had deployed OpenShift Container Platform, and our guess is that it is probably somewhere around 3,000 or so. We know that at the end of 2022, IBM said that OpenShift had an annualized run rate in excess of $1 billion. We are not sure if that includes the requisite licenses of OpenStack and Red Hat Enterprise Linux that are part of an OpenShift stack. Presumably so.
No matter. What is true is that the Red Hat business is still growing, but that growth has slowed. It is nowhere near the 20 percent growth that Red Hat was experiencing as IBM acquired it four years ago; Red Hat grew at 8 percent in the first quarter of 2023, to $1.52 billion. The growth is slowing because on a trailing twelve month basis is was up 11 percent to $6.17 billion. There is no reason why this OpenShift business cannot be many multiples larger, and there is no reason why a substantial portion of the 120,000-strong IBM i base and the 40,000-strong AIX base can’t be part of the OpenShift transformation. Even though many shops are not looking to install Linux.
OpenShift on Power has more than 250 customers, according to Steve Sibley, vice president and global offering manager of Power Systems at IBM, who gave out some stats at the POWERUp 2023 user group meeting last week, which we reported on here. It is not clear if even one of these customers was running IBM i on some of its Power Systems iron, or within logical partitions on its production machines, but probably not.
Linux is the preferred platform for so many data analytics and AI workloads, so we want to make it possible for IBM i shops to use these and easily integrate them with native IBM i applications. And I know a lot of you are skeptical and think you don’t need it, but if you want to run application development on a Web interface instead of a big fat client like Rational Developer Studio for i, the Merlin tool that was announced last year runs at least partially on OpenShift containers on Power. It is already PLEASE’d, as it were.
There can be hundreds of applications that run in the PLEASE environment, including HPC simulation and modeling and design software at manufacturers, route optimization for distributors, as well as AI inference and training workloads, which all pair very nicely with a Power10 processor. And we think there can be somewhere between 20,000 to 30,000 IBM i shops that could use this hypothetical PLEASE environment, and drive revenue for OpenShift as well as drive innovation for IBM i shops.
And to be honest, PLEASE is about the only way that IBM i shops are going to ever use OpenShift and Linux – just like PASE was about the only way that OS/400 and IBM i shops were ever going to use AIX was to use PASE.
Time To Design – And Deliver – The Application System/360
Historical, Functional, And Relevant
IBM Brings OpenShift Cluster Management Native On Power Iron
OpenShift Provides One Path To IBM i Modernization
Various Power Systems Updates And Tweaks
IBM Keeps OpenShift Up To Speed On Power Systems
Big Blue Rolls Out Red Hat Power Stack
Pushing The Capacity Envelopes With IBM i 7.3
Native Open Source: Why It’s Time for IBM i
A Different View Of IBM i And PASE
The truth can be painful for some! PLEASE tell me more moving forward.
The IBMi subsystem concept can be use for many purposes… in container style. Just let the system run also linux binaries beside aix ones or just implement the needed extensions on aix and just use the openshift thing over PASE++
Then just sell the Merlin thing directly to the IBMi shops to be run on i over “PLEASE” with all the necessary components, who cares if it works, it is easy for IBM, preserve economy, is supported, and bring peace to the world.
Just a note: PASE tools are straight AIX machine binaries, not TIMI stuff, maybe somebody cares as a tech difference if writes customer code.
bmw the IBMi can support even “functional serverless” style of computing. Think the *PGM as a function, that can even be replaced live.