The State Of The IBM i Base 2022, Part One: The Operating System
February 14, 2022 Timothy Prickett Morgan
You probably already figured this out, but we love data here at The Four Hundred, and we are always particularly keen on any data that helps us to understand the IBM i ecosystem better. And that is why we always look forward to the annual IBM i Marketplace Survey report that comes out in January each year from HelpSystems, and because we are thankful for the data they gather, we help push the survey every year and also participate in the webinar for it as well as a donation to the IBM i community.
IT Jungle co-editor Alex Woodie and I have been going over the data in the report and have done a few stories based on its findings, and the top three items in the Related Stories section below are derived from the current 2022 report, and a fourth comes from comments made by IBM during the webinar about the nature of IBM i and Power Systems tech support in the wake of the Kyndryl spinout. We will be extracting more stories from the report in the coming weeks, and this one you are reading is the next in the series.
Specifically, I plan to do a multi-part series on the state of the IBM i base based on the current report as well as historical data that goes all the way back to 2015, which is eight years of coverage encompassed by all of the reports. I want to start with the installed base of IBM i operating system releases as it has evolved over time. Let’s start with the data from the 2022 report:
This column chart is great for seeing how the distribution of operating systems changes within a year and if you squint a bit and focus on only one color, you can pick out the trends for each IBM i release. But it is hard to see the trends. Moreover, this data only goes back to 2018, and so I dug around on the disk drive to find the old reports and decided to not only add in the data for 2015, 2016, and 2017, but also to try to get a more complete picture of the waves of change in IBM i shops over time when it comes to OS levels.
Before I show this to you, though, I want to remind you of two things.
First, this question asks poll takers to pick what release level they are at for their primary IBM i operating system. It does not ask customers to pick all of their current operating systems by release and count, and it does not tell them to count across all of the machines at their main site and any remote sites that they may operate. Many customers have more than one Power Systems machine running the IBM i operating system, and they are not necessarily running all the same release levels on all of their iron. (This may not even be technically possible given the varying vintage of the machinery they might have installed.)
And second, as I point out every year when this survey rolls around, none of this data is necessarily reflective of the whole base of IBM i shops, of which there are around 120,000 unique customers worldwide. We all know that there are many, many shops running IBM i 6.1 and earlier releases, and this data has never really reflected that. Thus, we figure that the same people who respond to surveys and who read IBM i newsletters are also the same people who tend to be on relatively recent IBM i releases and Power Systems iron, who pay for IBM technical support for hardware and software. We call these the active users in the IBM i ecosystem, and we figure that it comprises around 30,000 customers. The remaining 90,000 or so IBM i shops are just as dependent on their systems as those who stay current, but for whatever reason – stubbornness, budget, or software compatibility issues – they do not stay as current. And some are downright laggards.
Inasmuch as I love my 2008 Dodge Dakota Sport pickup with the V8 Hemi engine in it, I get it. And I will keep that truck going for another decade if I can do it, maybe because we have seen a lot of the road together.
With that said, here is a line chart that shows the longer trends with OS/400, i5/OS, and IBM i releases:
In the more current surveys and reports, HelpSystems has lumped the older operating system releases into one category as new IBM i releases come out, which is natural enough. But we lose some visibility into what is going on with those older releases. So we peeled back through the old data and built a model to estimate the prevalence of the older releases among the active IBM i shops – those 30,000 or so companies we referenced above.
Some interesting things to note. First, there is very little of the older OS/400, i5/OS, and IBM i being used as the primary operating system in the active part of the IBM i installed base, and it is shrinking pretty much to zero at least among these 30,000 companies. Which is good.
You will also note that the rise of IBM i 7.4 counterbalances the decline of IBM i 7.2, and the rise of IBM i 7.3, which is still rising ever so slowly counterbalances the decline of IBM i 7.1. We know this is instinctively true because this is how a lot of customers tell us that they behave. They upgrade every other generation, skipping one in between because it is such a hassle to recertify applications. This makes perfect sense.
But as I say, there is a whole part of the OS/400, i5/OS, and IBM i base that doesn’t behave like this and that just sits there, on old iron and on older operating systems. And data like this above distorts how well or poorly we think we are doing. As I have said, there is a kind of System/3X base that is stuck in the past or stuck in the mud or whatever metaphor you want to use. And no one knows what this is doing to the base, and what opportunity it presents to modernize the IBM i base.
And frankly, I am sick of not knowing. And so I did some estimating of my own. Let’s assume that this other two-thirds of the IBM i installed base is indeed stuck in the past, and that past is around seven to ten years ago. If we apply the distribution of IBM i operating systems from the 2015 data to these 90,000 sites, that will give us a pretty good indication of where these shops might be today, I think. And if you do that, then this is what the installed base might look like overall, with the active and dependent shops added together:
Remember this is not the distribution of operating systems on all machines in the base, but just the primary production machine; and it says nothing about the operating systems installed on logical partitions. There might only be 120,000 unique IBM i customers, but they probably have an average of a between three and four machines across the base and well over 1 million logical partitions, all running an instance of IBM i, AIX, or Linux – mostly IBM i.
For those of you who like data in visual format, here is what I think the primary operating system distribution looks like:
When I speak on the IBM i Marketplace Survey webinars each year, I always assure everyone that the distribution of ages on the IBM i operating systems follows a pattern that is familiar with Windows Server and Linux versions and releases. All of these programs have seven years of standard support and three years of extended support, just like IBM i, and the behavior of enterprise customers is not all that different when everything is mission critical. But I am not so sure that the dependent but relatively inactive part of the IBM i ecosystem is like the Windows Server and Linux bases.
IBM i 7.4 is akin to Windows Server 2019. IBM i 7.3 is the same vintage of Windows Server 2016, which is now in year seven and heading for extended support. Halfway between IBM i 7.3 and IBM i 7.2 came Windows Server 2012, which is a decade in the field. IBM i 7.2 is like Windows Server 2012 R2, I guess. There is no good Windows Server analog for IBM i 7.1, but IBM i 6.1 (originally known as i5/OS V6R1) is coterminous with Windows Server 2008. OS/400 V5R3 is like Windows Server 2003 and i5/OS 5.1 (originally known as OS/400 V5R4) is like Windows Server 2003 R2. The alignments in time between these two server platforms is uncanny, really.
If this model of the primary operating system of the entire installed base is correct, do you really think that Windows Server 2008 and Windows Server 2012 and some even older Windows Server 2000 and Windows Server 2003 represent three quarters of the Windows Server base? I don’t. Maybe IBM i is more secure and Power iron is more reliable, and the workloads don’t change much and therefore it is easier for IBM i to hang back.
This is one of those times when I hope that my model is wrong. But I don’t think it is, and all of us have to start thinking, therefore, of this as an opportunity. The question is: How do we best attack that opportunity?
Maybe the Log4j vulnerability that is causing heartburn for many pieces of systems software will convince some customers to move ahead, but I have my doubts this will be enough. There have been no exploits for the Log4j vulnerabilities yet, and IBM is not going to patch the older releases to plug the security holes. That should give anyone running these vintage IBM i operating systems a cold shiver.