The Feeds And Speeds Of The IBM i Base
March 13, 2017 Timothy Prickett Morgan
Without change, nothing new happens. But change is always difficult, so there is always this push and pull between the way we have always done it and the way we might improve it, and there is also a kind of gravity associated with capital investments in hardware, software, people, and processes that impedes change – and often for the best. Otherwise, we would be changing everything all the time and no one would know what state anything is in.
The very stability of the IBM i platform, as measured in the way it runs and supports the applications on it as well as in the predictable improvements in the platform that come out of IBM and the third-party software (and sometimes hardware) vendors that create products for it, are both an achievement that is quite possibly unique in IT history as well as the very thing that keeps it from being like other platforms that have evolved along separate tracks and that are now more mainstream. This platform has been different since the days of the System/38, and is still different nearly 40 years later.
Back in the early days of the AS/400 three decades ago, we had access to a lot more information about that business and its customers than we have today. In the early 1990s, when we did our own estimates, we calculated that the AS/400 ecosystem generated something like $15 billion in sales for IBM and its partners, and by the middle 2000s, that had been more than cut in half as companies moved to other platforms. But there are still, depending on who you ask, somewhere between 120,000 and 150,000 customers out there in the IBM i base, and their ecosystem is still possibly generating billions of dollars per year in aggregate revenues and many orders of magnitude more in value for the customers who still employ the platform for their mission critical applications.
Who are these customers, and what are they up to now? Some of the best data that we have seen in the past several months came from the 2017 IBM i Marketplace Survey, put together by third party software maker HelpSystems, which is one of the largest players in the market. (It is not clear if Vision Solutions is bigger or not, since both are privately held, but both probably have sales north of $100 million a year.) We helped drive that survey, as we have for the past three years, and we participated with executives from HelpSystems and IBM in talking about the results of the survey a while back. But not everyone has time to listen in on a webinar, and I thought I would share my thoughts on some of the results of the survey that related to the demographics of the base. In a separate article, I want to address the concerns that many of us have over companies leaving the IBM i fold.
OK, maybe it is just me, but I find it funny that 1.3 percent of the people who answered the survey said they had no Power System iron, and I just noticed that now even though I looked at this data a bunch of times when participating in the webinars relating to the survey. Presumably they use IBM i cloud capacity and felt compelled to still participate. . . . The fact that a little more than a quarter of the shops polled said they had one machine is no surprise at all, and that the rest had two or more machines also fits in with what we know anecdotally about AS/400 through IBM i shops. Many customers keep old machines around to do odd jobs, and others use secondary machines for high availability clustering or to do geographic distribution of applications and data or to break up their workloads across business lines or divisions. I would like to have seen a split of companies with three, four, or five machines broken separately, because with more than half the base having from two to five machines, you can’t really see why they do this. The very fact that we have had logical partitioning, in one form or another, for nearly two decades would seem to mitigate against this proliferation of iron. You would do two machines for local HA replication and a third for geo replication, and carve out developers on a partition of their own on the same iron, thereby driving up utilization. But, processing and storage capacity on bigger machines is higher than on smaller ones, so this compels some customers, in some situations, from just consolidating workloads on the fewest number of boxes possible.
That said, the number of partitions in use at IBM i sites is on the rise, as you can see from the chart above, even though a little more than a quarter still say they have only one partition and are, in effect, running it as a traditional midrange system rather than one that is virtualized.
As you can see from the chart above, older operating systems are falling off the map, at least among that portion of the base that is active enough to read publications dedicated to the IBM i and respond to surveys. I like to remind everyone when I see data like this that we have a vast portion of the AS/400 through IBM i installed base that probably does not know it has one of these platforms running in their datacenters, and they just don’t think about it. They see the applications and that is what they know.
In any event, the older OS/400 5.3 and 5.4 releases and now even IBM i 6.1 releases are on the wane, and it won’t be long before they don’t appear in data even if they do persist out there in that invisible portion of the installed base for many years to come. It also looks like IBM i 7.1, which is many years old, has peaked, and IBM i 7.2 is on the rise and the shiny new IBM i 7.3 is starting to get some traction. This wave of upgrades looks to me to be more like the normal progression of operating system upgrades that I saw during the early days of AS/400, when IBM upgraded the hardware and software on an annual cadence, more or less, without making too many dramatic changes. Everything used to be incremental change, although some innovations were truly big deals. (I am not trying to minimize the past.) The important thing is that this looks normal, and the Technology Refresh pattern that IBM has established, with updates coming in the spring and fall and backcasting innovations into two different releases, is working to help companies consume the software in a predictable fashion and do a better job staying current.
This is a very good thing. It is a pity that upgrading hardware is not so easy – and we have often mused about the ways that IBM could make it so. That said, companies are still moving ahead, as you can see here:
Power5 machines date from 1995, if you can believe it, so they have been in field a long time, and those that are older than Power5 in the chart above could be truly ancient, as least given how a typical X86 server has a lifespan of three or four, and sometimes five, years. IBM i boxes go in and stay in for a lot longer, which is again a testament to the stability of the platform and the applications that run on them. The database, as we have said many times, is the stickiest thing in the datacenter.
What is immediately obvious from the chart above is that companies do not stay as current on hardware in precisely the same way as they do on software. Processors support multiple operating system versions and releases, but they eventually age out and have to be changed or stay in a kind of stasis. This is why more than half of the base taking the survey said they had Power7 iron (which could mean Power7 from 2010 or Power7+ from 2012) and only 40 percent said they were on Power8 iron (which made its debut in 2014). Eventually those Power5, Power5+, Power6, and Power6+ machines will be retired, but if you look at the rate of change, there is nothing compelling it to move very fast. Perhaps some wheeling and dealing by IBM with the new Power S812 Mini will help move things along. The IBM i license transfer deal announced recently will help, but it is probably not enough.
IBM i shops used to be fairly monolithic back in the early days, but datacenters started getting more diverse as file serving and then Web applications came into the fore. Mergers and acquisitions over the ensuing decades have also made datacenters more diverse in terms of their mission critical platforms, and of course, Power and X86 and other kinds of systems have virtualization technologies that allow for multiple, concurrent, and sometimes incompatible operating systems to share a single physical server. And so, operating systems proliferate, as you can see here:
The interesting bit for me in this chart is that the distribution of Linux on X86 iron and Windows on X86 iron matches the world at large. For the longest time, Windows dominated X86 iron installed in the OS/400 through IBM i base. It won’t be long before Windows and Linux are at a 50-50 split, I think, because Linux is the only operating system on the rise. And by the way, Linux on Power is on the rise, and partially for the same reason. All other operating systems are flat or in decline.
It is also interesting to note that a little more than a quarter of the IBM i shops out there have only IBM i in their datacenter. They are old school in their style, indeed. And if it works and they are happy, why not?
This last chart from the survey talks about applications, and it is finally, I think, a set of data that is reflective of the reality that I knew was out there and that reflects the heritage of the AS/400 platform. Take a look:
There are lots of companies using third party applications, to be sure, and we have never contested this. But the fact that nearly two-thirds of the shops polled said they write their own in-house applications demonstrates what we all knew all along. So if you are using code that might have come from a third-party vendor one, two, three, or even for decades ago, or if you helped create whole new applications from scratch, the data now reflects you. There are several hundred thousand of you RPG and COBOL and PHP programmers out there, encapsulating your company in your code. We honor you and your efforts, and we thank you for reading The Four Hundred.