Two Top i Concerns and a Bunch of Little Ones
June 14, 2010 Timothy Prickett Morgan
Our colleagues at COMMON Europe will this week present the first pass on their annual Top Concerns survey, which the IBM midrange user group has put together for the past five years to help steer Big Blue’s marketing and development of machines running the i and predecessor operating systems and their RPG language and integrated database. I have no doubt that many users will give IBM some good advice on what it needs to do. As an advocate for i shops, I have two toppers of my own and a bunch of little ones that I would like to toss into the pile.
Feel free to send me your thoughts after reading this. I will be happy to pass them on to readers of The Four Hundred as feedback and to make sure that Ross Mauri, general manager of the Power Systems division, sees them.
Topper Number One: The OS/400 and i5/OS installed base running on vintage or aging servers is not moving ahead fast enough.
This is the curse of designing a rock-solid system that has long-term support and that is typically used to support workloads at small and medium businesses that have steady or modestly growing workloads. Windows and Linux are still growing up, so you have to move to keep pace, much as AS/400 shops had to do from 1988 through around 1993 or so. If you wanted more performance, you had to keep pace with the software, and the features being added to OS/400 throughout these five years were so valuable you were chomping at the bit to keep moving. And this was despite the fact that processing capacity was ridiculously expensive; the software was relatively cheap by comparison, and IBM was pretty generous with it.
There are many things that could be done to help get AS/400, iSeries, and System i shops to jump from where they are all the way up to i 7.1. And I think the answer is to reach back in history and hit repeat. What IBM needs to do is to admit that anything that predates a Power5+ at this point is effectively a System/36 in 1993: a dead account that needs to move forward. Back then, after years and years of failed attempts to move System/3X customers forward, IBM threw its hands up in the air and ported SSP, the System/36’s operating system to a funky new 64-bit PowerPC processor and called the new box the Advanced/36. (Remember, this machine got IBM’s first 64-bit processor in 1994, well ahead of the formal launch of the PowerPC chips in the AS/400s in the summer of 1995.)
So IBM needs to create a little something I will call the AS Emulation Environment, an analog to the S36EE runtime for RPG II applications that came embedded in the original OS/400. The S36EE ran with degraded performance, to be sure, and IBM caught hell for it at the time, as you well remember, but these days, Power7-based machines have more raw power than i shops know what to do with. I say burn those clocks and do an emulation mode for the OS/400 V5R3 level of the platform on these new boxes and completely insulate these applications and their users from i 7.1. If need be, use that QuickTransit emulation software and embed it inside of a logical partition, but get it done. Charge nothing for this feature, and make sure it runs on the appropriate machines to attract that installed base of customers who are still on OS/400 V5R3 or i5/OS V5R4–the former which is dead and the latter which should be, but IBM had to extend its life last fall because people do not want to do program conversion no matter how easy IBM makes it sound.
If Apple could avoid program conversion by using QuickTransit as it moved from PowerPC to X64 chips, I fail to see why the same approach can’t work with older OS/400 and i5/OS applications. Get these people on new iron, and then gradually move them over.
Topper Number Two: The i base needs inexpensive, flexible, inexpensive, powerful, inexpensive, and upgradeable entry Power Systems supporting i 7.1.
Did I say inexpensive? I’m just checking to make sure.
Look. No matter how good IBM says the yields are on the Power7 chips, there are a slew of them ending up on the scrap heap that have more oomph, even with most of their cores and cache memory covered in semiconductor boogers. And these throw-away chips can be the foundation of a new AS/400 business.
Here’s the way to think about it. Let’s say that no matter what Big Blue does, the Power Systems-i business is only going to generate $2 billion a year in revenues for the company. Now, if everything is priced somewhere between the Unix and mainframe market, but the customers are working on Windows budgets, IBM will be able to get a grumpy percentage of customers with legacy applications to begrudgingly pay for the capacity they need while everyone else will hold back, getting by and trying to figure out how to never put new applications on the box.
There’s another way to get that same $2 billion, and I am sorry, IBM, but there is no way to easily make it $5 billion, but if you are lucky, maybe it can be $2.5 billion. And that is to cut the customer base the kind of deal they ain’t seen since 1995. Or better. And then have the whole customer base get current at the same time. This is what all of the X64 vendors are trying to do right now with aggressive pricing on the latest Xeon and Opteron boxes. They are praying customers just start throwing out all the old stuff and they are most definitely going to try to make it up in volume. But they know that this will not get them anywhere near the peaks of the dot-com boom revenue wise.
So you can make customers happy and make $2 billion or make them miserable and make $2 billion. Happy is always smarter than the miserable route because that customer remembers you cut them a break and maybe, just maybe, they might put a lot more work on the boxes and buy lots more capacity.
Other relatively minor Top Concerns: Here’s a few more relatively minor ones, in no particular order.
We need a cheap development machine. As I have said before, it doesn’t matter if the whole environment is running in emulation mode atop Windows 7 or Linux.
We need live migration of running workloads across multiple machines. How embarrassing is it that OS/400, which pioneered logical partitioning in the midrange in 1998, has been surpassed by all the major X64 hypervisors and AIX in having live migration? Quite, I think. Live migration allows for teleportation of running workloads from one physical machine to another for maintenance and disaster recovery purposes. Yes, this will eat into high availability software sales. Yes, you can work with HA partners to have them be part of the plan. But this needs to get done.
We need proper co-processors, like we used to have. The server world is coming around to the idea of having integrated co-processors like the AS/400 had starting in the early 1990s. And yet the Power Systems-i machine is moving in the opposite direction. You had the right idea. Put an integrated X64 server card back on the bus and forget trying to peddle BladeCenter. It is wicked overkill for i shops. You had the right idea to begin with. Put a server on an X64 bus, get a Xen or KVM hypervisor to abstract it so Windows or Linux operating systems don’t know, and sell a bunch of these and make i shops happy.
We need a product name that gives this machine some dignity. Ask any customer, IBM. Ask every customer. I’ll make you a deal: Call it the I Business Machine, with a capital I, and we’ll call it even.
P.S. Whatever happened to the Smart Cube appliances? There were some good ideas in that box. Put them into the generic Power Systems lineup and make remote support like we all know IBM can do sit up and bark.