The State Of The IBM i Installed Base, Part 1
February 10, 2020 Timothy Prickett Morgan
This is the time of the year that people assess where we are at so we can figure out where we are going. So we think it is appropriate to ponder the current state of the IBM i business.
There isn’t just historical precedent for pondering the state of things this time of year; the pondering, in a certain way, drives history. The end of the prior year is mostly gone from memory and focus shifts toward fulfilling the promise of the current year. It is human nature to mark time this way, driven in part by the seasonal nature of our agrarian roots. When you are harvesting, you don’t have time to think, and in the dark days of winter, you are just trying to survive and get to the holidays. It is in the late winter and early spring that you prepare for the next season of planting – or for the war to protect the realm. By fall, the wars, if they come at all in a given year, are over, and you go home to harvest and pick up the pieces. Sometimes we might go a decade without a war or a recession, which eases the burden to be sure, but does not change the cycle.
IBM does not supply a lot of data about the Power Systems hardware platform and its IBM i software platform, so we have to gather up whatever data we can find to try to figure out where we are at and where we might be going. In the past six years, thanks in part to the encouragement of the late Dan Burger, our executive managing editor for nearly two decades here at IT Jungle, IBM i systems software provider HelpSystems has been doing a survey of the IBM i market and we now at least have some kind of dataset with which to monitor some trends.
The sixth annual IBM i Marketplace Survey report was released by HelpSystems in mid-January, and it was based on a survey that had over 500 respondents, which is a pretty good number. I participated in the webcast going over the survey results along with people from HelpSystems and IBM, as I have done for the past several years. Tom Huntington, the executive vice president of technical solutions at HelpSystems, hosted the webinar last Thursday, and we were joined by Alison Butterill, IBM i product offering manager, and Ian Jarman, who had that job before and is now an executive in the IBM Lab Services division, and Brandon Pederson, who is worldwide IBM i product marketing manager. You can get a copy of the survey report and also listen to a replay of the webcast at this link. We have been interpreting the report results in a series of stories, and we already covered the plans that people have to upgrade their Power Systems hardware and IBM i software in 2020. This week we are looking at the machine distribution and logical partition use among the IBM i base, and in a follow-on story we will see how the Power9 ramp and IBM i 7.3 and 7.4 ramps look compared to prior generations of technology.
Let’s start with the demographics of the systems first, and we will start with these two charts that show the distribution of machines by number and by number of partitions (we presume total partitions at the site and not per machine) at the customers who responded to the survey in late 2019 for the 2020 report. Take a look:
Historically, long before there was logical partitioning available on systems, companies had a big machine for their transaction processing workloads, large enough to meet their peak demand at end of week, end of month, and end of year and whatever seasonal spikes they might have in processing; application development and testing was usually done on a second machine, so the average across the AS/400 base back in the day was a little north of two. But over the years, with the advent of logical partitioning, dev/test workloads could move onto the production box, reducing the system count, but at the same time high availability clustering within the datacenter and disaster recovery replication to remote datacenters also increased. There has also been a certain amount of consolidation in industries across nations, which often need distinct systems for each country or region they operate in, and thus, the average number of machines per customer has been on the rise. It could also be that the one machine, Mom and Pop shops have left the base, allowing the average to rise as well. I suspect that all of these factors have come into play.
I am not one to just leave a dataset alone; I have to play around with it to have it make sense to me and to see if what it is telling me rings true. So I did a little math on the raw data gathered for the most recent IBM i Marketplace Survey 2020 report. Take a look:
We know that “over 500 people” responded to the survey, and we got the exact number of 506 from HelpSystems. (Thanks for that, Tom.) I took the distribution of customers with a particular number of machines and multiplied that number by the number of customers to calculate the total number of machines in the distribution by the tiers that HelpSystems set. I took the midpoint of each tier when there was a range, as shown in the table, and for the 31+ category, I just set it a little higher than 31, at an average of 42 machines per company. This is an average that makes sense to me, given that there are only 180 countries with modern economies in the world, and there are very few companies that play in all of them. (IBM is, in fact, one of the few that does.) It is admittedly a guess, but no matter how you change the number, it doesn’t change the overall result that much.
If you do this math, what you see is that when it comes to machine count, companies with four or more machines utterly dominate the base even as those with three or fewer machines dominate the company count. This is an important distinction. What those numbers in the table above show is that while customers with three or fewer machines account for 75 percent of the machines at these 506 sites polled for the survey and report, they only account for 29 percent of the machines in the base. We always knew that companies using big AS/400 and IBM i hardware dominated the revenue stream for IBM, but what this also shows is that companies with large numbers of IBM i systems also tend to dominate the installed base. This suggests that single-system shops do not matter as much to IBM’s plans as those who have big iron or lots of small, medium, or large iron scattered around their companies.
This is a subtle distinction, but an important one. There may be, depending on who you ask, anywhere from 120,000 to 150,000 unique IBM i shops in the world, but those with onesies and modest workloads are not driving the revenues.
Now, given the average distribution of machines at 4.48 per customer (not per machine or per site, but per customer) in the sample of respondents to the IBM i Marketplace Survey, this implies that if there are 120,000 customers worldwide, then there is an installed base of just over 535,000 machines worldwide in the IBM i and predecessor base, and of there are 150,000 customers, then the base numbers just under 670,000 machines. I do not think this is true, but I am happy to be wrong.
One think that this survey did not explicitly ask – and we do not know what respondents are counting – is if they have hosted IBM i iron in the cloud. It would be interesting to see that counted out separately. While we think this number is small now, it is no doubt on the rise and will rise further. The same holds true of cloudy partitions on public clouds where customers buy a slice of a machine rather than a whole machine. These will need to be counted in the partition counts going forward, obviously.
The distribution of logical partitions across the last three years of the survey is shown in the first chart in this story, and as before, we did a little math and analysis on this. Take a look at this table:
There are obviously a lot more partitions in the IBM midrange installed base than physical machines. We did the math on the distribution once again as we did for the machine count, and found that the average customer had 13.5 logical partitions in total, and that this pool of 506 survey respondents had what is probably shy of 6,900 LPARs across its base of around 2,270 physical machines. (There is probably 5 percent wiggle here and there in these numbers, maybe a little less.) The important thing is to once again see the distribution of partition tiers by customer count and then actually count the partitions at these customer tiers. In this case, 68 percent of customers have six or fewer partitions and the remaining 32 percent of customers have more than this. But if you do the math and count the partitions, then only 13 percent of the partitions are at customers with six or fewer LPARs and the remaining 87 percent are at customers with seven or more LPARs. In fact, those with between 11 and 30 partitions account for 25 percent of the aggregate LPARs in this sample and, assuming the average is around 150 for those with 101 or more partitions (the highest tier), then a third of the base has a large number of partitions spread across those machines.
Now here is the fun bit. If this distribution of logical partitions is across these customers is correct, then that implies the base of 120,000 customers would have an aggregate of 1.6 million partitions and the base of 150,000 customers has over 2 million partitions. Let me make this clear: 2 million partitions is a lot of partitions, for any customer base. But, given that OS/400 and IBM i have had logical partitions since OS/400 V4R4 back in 1998, I am not surprised that the technology is well established. My guess is that across the total installed base, not the sample in the survey, the customers look more like the bottom piece of the base than the top, so the average machine count and partition count is lower. I have no problem believing that there are 250,000 to 300,000 machines in the base and maybe somewhere between 1 million and 1.5 million partitions. As I have pointed out many times before, I think the people who take surveys about the IBM i platform are the ones who know and care about the platform and who are more active in their development and in keeping their hardware and software current. So of course they are doing more modern things.
In the next part of the analysis of the installed base, I will look at the currency of the hardware and operating systems among those surveyed.