The Windows Of Opportunity

November 17, 2014 Timothy Prickett Morgan

It has been a long, long time since the AS/400 was the dominant, and some cases the sole, platform in use at most IBM midrange shops. As we all know, the client/server revolution took the data center by storm in the late 1980s and early 1990s, first on the desktop with the PC being used as a kind of intelligent monitor for back office systems and then later with the popularity of NetWare, Unix, and eventually Windows on servers, in the data center itself as adjunct coprocessors for systems like the AS/400.

Because Windows was familiar and then ubiquitous on the desktop, it became the natural server environment for the many millions of small businesses that started out on a PC and then grew into a data closet, a data room, and maybe eventually a full-blown data center. Even when large companies used AS/400s and their progeny for key, mission critical applications–as Microsoft itself did for many decades, you will remember–those applications were surrounded by a wall of X86 servers that were used to provide gateway services into the AS/400 applications for the rest of the network as well as to run additional applications that culled data from the OS/400 machines or that did their own work, such as running email and collaboration servers. IBM made a stab at bringing such front office applications to life on the AS/400 with its OfficeVision/400 suite and then in the mid-1990s bought Lotus Development Corp to counter the collaboration software that Microsoft, Novell, and others were peddling just as the Internet was taking off as yet another application revolution.

Now, two decades have passed, and the AS/400 platform has evolved into Power Systems machines running the IBM i platform, but the hybrid architecture of the AS/400 data center remains about the same. Most IBM i shops run a mix of IBM i and Windows workloads, with NetWare and Unix pretty much vanquished from their data centers and Linux not really taking off as IBM had hoped either on the Power platform or on X86 iron. The last time I could any good statistics out of Big Blue about this indicated that more than 85 percent of OS/400 and IBM i shops had Windows platforms running side-by-side with these machines, and the indications are that if they had one or perhaps two and maybe three Power-based systems, they had several multiples more of X86 iron.

If the behavior of these customers is consistent, then it stands to reason that a minority of these customers keep current on either IBM i or Windows. From the data I have seen and the models I have created based on it, it looks like there might be something on the order of 30,000 OS/400 and IBM i shops that keep current out of a base that IBM says is at least 150,000. Assuming that the average customer has one machine but some have multiples, then there could be as many as 200,000 systems out there, many of them on Power5+ or earlier iron. And, there could easily be anywhere from several hundred thousand to a few million X86 servers, most of them running Windows, wrapped around them or sitting nearby.

The launch of the Power8 machines, IBM’s sale of its System x server division to Lenovo Group, and the expiration of support on Windows Server 2003 in September 2015 presents Big Blue with a unique, once-in-a-decade opportunity. Depending on who you ask, there are somewhere between 8 million and 12 million Windows Server 2003 machines out there in the world, all of which need to be upgraded because no one wants to leave a Windows server unpatched. The thing is, the move from Windows Server 2003 to Windows Server 2012 is going to be extremely disruptive, and Windows Server 10 (if the next release of Microsoft’s server platform is indeed to be called that) is just around the corner, too. IT shops using this older version of Windows, whether they have OS/400 or IBM i or not, are going to have to go through a bit of pain upgrading their base Windows platform and either porting their applications over or getting new applications.

Any time a company is looking at making such a big jump, it is an opportunity for a new platform to get its foot into the door. Now that IBM does not have a System x or Windows business to protect, it would be a good time for Big Blue to come up with some kind of strategy to try to pull as many of those workloads off of Windows and onto equivalent software stacks running on its Power Systems. Because you can sure bet that IBM’s X86 rivals such as Hewlett-Packard, Dell, and now Lenovo are licking their lips over the opportunity to go into midrange shops with vintage Windows systems and convince them to not only upgrade those Windows machines to current software, but also to move as many applications as possible off of OS/400 and IBM i systems at the same time.

As I have pointed out many times before, this all would have been a lot easier if IBM and Microsoft had kept the Power version of Windows Server alive and not killed it off before it even got going. But like IBM, Microsoft was conflicted about having many different chip architectures to support, and frankly has been happy to drop support for MIPS, Power, and Itanium architectures, and it is not exactly pleased at the prospect of having to consider adding ARM processor support for Windows Server should that part of the market take off. The difference now, of course, is that Microsoft is a hyperscale data center operator and it needs any edge that any architecture can give it with its more than 1 million server farm. Microsoft might need ARM for its own uses so badly that it does the Windows Server port for its Azure cloud and platform services because it wants or needs tighter control of its hardware stack and lower-cost options.

IBM has a few options as the new year gets rolling to try to capitalize on this window of opportunity. The first is to simply help customers who have vintage OS/400 and Windows platforms upgrade those platforms in lockstep. IBM might not sell System x iron any more, but it certainly can sell services to help customers do upgrades and it certainly can work out a deal with Microsoft and Lenovo to spearhead the upgrade efforts. IBM could do a deal that gives customers special price breaks or other incentives if they upgrade to IBM i and Windows Server 2012 at the same time.

Another approach that would probably be a lot easier to pitch to the top brass at IBM would be to help customers running either vintage IBM i or AIX releases to upgrade to the latest IBM i 7.2 and AIX 7.1 operating systems while at the same time moving as many Windows workloads over to their Linux equivalents on Power Systems iron. This is not an easy thing to do, but it is not impossible. A lot of the Windows servers at IBM i shops are running Exchange Server for messaging and groupware or SQL Server for OLAP serving and data marts. All of these functions can be ported over to Linux on Power, and if IBM spent a little money, it could do so in a way that minimizes the disruption to Windows client users on the desktop. For instance, OpenExchange Server does a pretty good job of faking out Outlook that it is in fact Exchange Server. IBM’s DB2 database for Unix and Linux can “skin” an Oracle database, but as far as I know it cannot make itself look like a SQL Server database. This is precisely the functionality that might get IBM i shops to contemplate moving Windows database workloads onto Power partitions.

I realize how difficult it will be able to get IBM i shops with Windows workloads to move SQL Server databases onto DB2 for i. Even if Big Blue gave the IBM i operating system and its integrated DB2 for i database away, customers would be hesitant because the databases are so different. But if it drives Power Systems sales and porting such workloads is easier than I think it is–it would be nice to be wrong here–then even giving away free IBM i licenses might make sense. So, for instance, IBM could say that for every customer on a vintage iSeries or System i machine that ports over a core onto a Power8 box running IBM i 7.2, they get a free core, complete with an IBM i license, to port over work that was formerly done on SQL Server. Or maybe it is two cores free for every one they pay for. Ditto for Exchange Server and other Microsoft middleware and systems software.

The point is, there is an opportunity here and IBM has to jump at it. Such an opportunity may not present itself again for a long time. The question is whether IBM has the services skills and the will to make such an aggressive push. If IBM creates the program and involves partners, such a push to migrate customers off Windows and onto Power could be much more important for the long-term viability of the Power Systems line than any new generation of chips.