The IBM i Base Did Indeed Move On Up
January 21, 2019 Timothy Prickett Morgan
This time last year, in the wake of the 2018 IBM i Marketplace Survey report put together by HelpSystems and based on survey data gathered in the fall of 2017, we said that the IBM i base was ready to move on up to newer iron. And guess what? Based on the results of the survey done in October 2018 and released in the 2019 IBM i Marketplace Survey unveiled last week, it looks like a pretty healthy portion of the base did in fact get off older iron and move to shiny new Power9 iron. In some cases where customers have to stay back on IBM i releases due to their application software, some shops probably moved to Power7+ and Power8 iron that was newer than what they had, too.
The IBM i Marketplace Survey was conceived of more than five years ago by Tom Huntington, the executive vice president of technical solutions at HelpSystems and a Power Champion for as many years now, and the late Dan Burger, executive managing editor here at IT Jungle until he passed away last summer. Back in 2014, when Huntington and Burger were kicking the idea around, there was not a lot of demographic information about the IBM i installed base – Big Blue had long since stopped giving out even the meagerest of information, and the big market researchers like IDC and Gartner had similarly not given the IBM i platform its due. The idea was to use the remaining IBM i publications to get readers using the platform to tell us what they were up to in terms of hardware and software and what their concerns were in a number of key areas such as security, application modernization, and budgeting. And, five survey reports later, we can say that this has been a successful endeavor indeed.
For this year’s report, the survey done in the fall had 700 respondents, with 57 percent coming from the United States but only 2 percent from Canada; another 19 percent came from Europe, 15 percent came from Latin America, and the remainder was split across Australia, Asia, and Africa. This is not precisely representative of the installed base, which is more heavily distributed outside of the United States, but the numbers are gradually shifting in the right direction as HelpSystems translates the survey into more languages. (It was available in English, French, Spanish, and German this year, and Japanese and Chinese should probably be added next.)
Over the next few weeks, we will take a dive into various aspects of the 2019 IBM i Marketplace Survey, but this week we are going to focus in on the vintage of Power Systems iron and IBM i operating system releases in the installed base.
The organization of the Power processors in the systems and the software releases on them is a lot more useful in the 2019 survey report since it is not organized by Power chip or OS release, but by year, and thus you can better see the trend lines in the bar charts over time for these different aspects of platforms in the installed base. When looking at this data, it is important to realize that many shops have multiple machines, and often multiple partitions on the systems, capable of supporting multiple versions and releases of operating systems, so the mix adds up to more than 100 percent for both hardware type and OS release. This data is not gathered per site, but per company, and many companies have multiple datacenters (to cover countries or regions), and many also have multiple machines, with the main production system plus a development system and maybe a backup/recovery system locally and perhaps even one in a remote location. The systems add up, and the OS levels are fruitful and multiply.
Here is the distribution of the number of Power systems installed at the companies that responded to the most recent survey:
After both Alison Butterill, IBM i product offering manager at Big Blue, and I crabbed that the prior surveys did not offer enough granularity between two and five systems, Huntington tweaked this year’s survey and now we can see who has one, two, or three machines. The None category is a bit perplexing in the chart above, which I just noticed as I write this. (And that is after participating in two live webcasts last week going over this data with Huntington, Butterill, and Ian Jarman, who used to be IBM i product offering manager and is now an executive in the IBM Lab Services division.) On first blush, None implies these companies should not even have taken the survey at all, but this is (presumably) for companies that rent rather than own their machines. Or, it really is for people who should have deferred from taking the survey. In any event, just under a third of the sites have only one machine, which means their entire company is running on one box and they do not have high availability clustering software from the likes of Maxava, Syncsort (which owns Vision Solutions and Trader’s), HelpSystems (which owns Bug Busters), and Rocket (which bought iCluster from IBM after it acquired it from DataMirror). Those with two machines could have a production and a development box, or HA clustering, and those with three machines could be doing both, or neither with three different regional or national datacenters.
As you can start seeing, this can quickly devolve into a lot of guessing. And when you throw in the partitions, of which there are many on IBM i systems, this can get very complex indeed – more complex than you can characterize from a survey that takes 20 minutes or so to complete. Here is the partition distribution:
What we can honestly say is that the number of partitions is low in some cases, and high in other cases but nowhere near the practical limits of the PowerVM server virtualization hypervisor. My guess is that about a third of the systems have only one partition (it is in essence a bare metal machine), and the rest is a mix of production, development, and regional boxes with a mix of partitions on those boxes. At this point, with logical partitioning being available for two decades with the OS/400, i5/OS, and IBM i operating systems, it seems logical that companies are using this technology to support workloads and isolate them and sometimes whole operating system instances with regional and national settings.
For this year’s report, the Power6 and Power6+ machines are now glommed into oldest group of hardware vintages, which including Power5+ and older iron last year. In this data, the so-called “plus” versions of the Power chips – Power5+, Power6+, and Power7+ – are put into the same generation as their predecessor full versions – Power5, Power6, and Power7. (Note: There was no Power8+, and there is not going to be a Power9+, although there are Power8′ and Power9′ machines that had I/O rather and core microarchitecture and/or chip manufacturing process improvements that are typical of the “plus” models.) Take a look at the chip version distribution from the current survey:
Here’s the interesting bit in the chart above as far as I am concerned. Back in 2014, after the Power8 chips had been shipping for a few months and the survey was launched for the 2015 report, only 8 percent of IBM i customers had moved up to Power8 iron. The Power8 chip has gradually been adopted over the ensuing four years, and accounted for 57 percent of the installed base as 2018 came to a close. The Power9 chips for the IBM i platform came out last February 2018, four months earlier in the year than the Power8 entered the market, and that in part has helped for the Power9 processors to have a higher market penetration relative to Power8. But more than that, I think that customers are just ready to move up to more modern machinery. And in fact, I think that a lot of shops with Power5, Power5+, Power6, and Power6+ machines moved up to second-hand Power7 and Power7+ machines as well as moving to Power8 machines a few years ago and Power9 machines last year. Power9 systems, as you can see above, are now installed by 14 percent of those who took the survey, which is considerably larger than the Power8 penetration five years ago. Power6+ and older systems are declining pretty fast and Power7/Power7+ installations, which have been pretty steady, look like they are starting their decline as Power9 machines are on the rise.
It is important to remember that this is survey data, taken by a self-selected group of customers who are generally more active – they read industry publications or promotional material from vendors – so this may not reflect the actual installed base, which I think has a lot more vintage iron than this distribution shows. But I do think that this distribution of iron does accurately reflect the customers who tend to stay somewhat current.
It is in many ways easier to stay current on operating systems than it is on hardware, since processors from one time can support multiple future IBM i releases that come out in the future. (This is true of all processors, not just Power Systems machines running IBM i.)
Among the active shops represented by those who took the survey over the years, IBM i 7.1 (which is now on extended support) is on the decline, IBM i 7.2 is holding steady (but there is something a little weird in the data in 2015 and 2016, with the share of IBM i 7.2 going down by half before rising to its current levels and holding there from the surveys in 2017 through 2019. IBM i 7.3 is clearly on the rise, with 42 percent of shops moving there as of last fall.
For IBM i shops, there is a new Power processor every three to four years and a new IBM i release every 18 months to 24 months. So, ironically, the software is on a Moore’s Law curve and the hardware is on a 2X Moore’s Law curve.
Now, here is the neat bit and a cause for optimism here in early 2019. Despite the fact that the world might be on the brink of self-inflicted recession (the U.S. government shutdown, trade wars between the U.S. and China, and the Brexit exit of Britain from the European Union) that could be lurking on the horizon, IBM i shops look like they are gearing up to upgrade their hardware and software – and sometimes both. Take a look:
Some 58 percent of those surveyed say they are going to upgrade their Power Systems hardware, their IBM i software, or both in 2019. And nearly a third, if you drill down, say they will be upgrading both. This is a very big leading indicator for the IBM i business, which had growth in the first three quarters of 2018 and which we hope continued to grow when IBM tells us all about it when it reports its fourth quarter financial results this week.