The Big Two Four For The Four Oh Oh

June 25, 2012 Timothy Prickett Morgan

When you are a computer system of a certain age, you have to count your blessings as well as your aches and pains. As I sit here on June 21, 2012, I just took a walk in Manhattan down on Broadway to intentionally experience the 97-degree heat that is still building toward 100 this afternoon. I didn’t take a walk because this heat wave would clear my head and somehow focus my mind, but rather because for once I would be forced to walk at a lazy pace and let my mind wander through all the many years we have spent on the OS/400 and IBM i platform together.

The same high pressure system that is baking New York City is the one that caused low pressure systems to back up over the Old Northwest, causing up to 10 inches of rain to drench northeast Minnesota, unleashing massive flooding on the St. Louis River that, thus far, has not directly killed anyone as the water tries to make its way downhill to Lake Superior. We live in extreme times, and not all clouds are computing utilities with silver linings and no man is an island, unless, of course, you have $36 billion like Oracle co-founder and CEO, Larry Ellison and you want to buy up Lanai, an 88,000-acre Hawaiian island, as your private playground for an estimated $600 million. Pocket change, and remember that the next time your Oracle software bill comes in.

We live in a world where most things that affect us are out of our control. This, of course, can make us uncomfortable, if not downright ornery.

For me, I try to focus on the work, on the things I can control and put my energy into and get some fulfillment out of. Some days, it is the work that drives us crazy, and some days, it is the work that keeps us from going insane. There’s nothing more dangerous in the world than somebody with no problems, with nothing to do but make trouble. And I am grateful that being the editor of The Four Hundred–with a several-year interlude writing Monday Morning Update at Midrange Computing–gave me something to do and that only sometimes got me into trouble. The good kind of trouble, like explaining what new technologies are available, how you might use them, what they cost to deploy, and how they could be improved.

A lot of what I learned about the world, I learned from writing this newsletter, and unlike a college education, this one actually paid me to do the studying. I learned about business and technology and how competitive and often unfair the market can be, how difficult it is for any new technology to establish itself, and how challenging it is for an established technology to continue to evolve. The story of the AS/400, no matter what you call it, is a story much like our own. It has a beginning, and a middle–you’d expect that from a midrange platform, after all–and it will no doubt have an end. It has a remarkable narrative arc that all of us participate in each day. We keep that story alive every time we do our work. There is honor in that, as well as a paycheck.

Personally, I am amazed that the IBM i platform still exists at all, and I am glad that it does for perfectly enlightened and self-interested reasons. Customers of other proprietary platforms, like Digital’s VAXen and their VMS operating system, and Hewlett-Packard‘s HP 3000’s and their MPE operating system and integrated database management system, have been treated more roughly than OS/400 and IBM i shops. Maybe that is enlightened self-interest on behalf of Big Blue, and maybe it is just that it doesn’t cost IBM that much to tweak IBM i and make 100,000-plus customers worldwide happy computing on their existing platforms. I can’t say I understand IBM’s motivations, without seeing the profit and loss statement for the product line.

But what is clear is that once IBM is committed to the Power chip for running AIX and Linux and IBM i workloads in the data center–as well as Windows in game consoles for Microsoft and other operating systems in consoles from Sony and Nintendo plus myriad embedded applications–the extra effort to keep IBM i alive is probably not all that great. But if IBM loses out on the next wave of exascale supercomputers using either BlueGene/Q or Power 775 clusters or loses the console chip business to Intel, Advanced Micro Devices, or any one of the ARM collective, then all bets are off. The fate of the IBM i platform is not in its own hands, but rather on many other businesses that IBM is engaged in. This can be a little unnerving, if you stop to think about it. But the next generation of chip fabrication plants to do 450 millimeter wafers and maybe 14 nanometer and 10 nanometer processes are going to cost billions and billions of dollars, and IBM is going to have to come up with the money somewhere. Or farm the work out to a fab China, as it already is building a lot of its servers–including the new Flex System nodes as well as Power Systems iron–today.

I think it is useful, on the 24th birthday of the AS/400, to think about how far this platform has come without sacrificing its principles. That didn’t happen by accident, but by design and through the perseverance of countless dedicated people inside of IBM and unsung hordes of those of us on the outside, who nudge IBM here and there to keep moving in new directions that the IT market is moving.

IBM was half the size it is today and largely dependent on hardware sales, with a smattering of software and services, when the AS/400 was launched. And at the time, IBM had about a quarter of the entry and midrange systems market, about $30 billion a year, peddling its System/36 and System/38 minis. The “Silverlake” project took three years to come together, and was key to IBM retaining its 275,000 or so midrange customers. There was no PC server back then, of course, unless you count Novell file servers, and IBM’s baby 9370 mainframe, promoted as a “VAX killer,” was panned because it didn’t really sell well at all against the VAX when it was announced two years earlier. This was a time when IBM CEO John Akers was trying to revamp Big Blue and make it more aggressive and nimble while it was still under the watchful eye of the Antitrust Division of the U.S. Justice Department. And as we know in hindsight, a few years into the future after a stock market crash and a recession, IBM almost went bankrupt.

This was not an easy time for anyone then, either, and it is helpful to remember this as we deal with our own issues here today.

Let’s take a look at the main engine in the AS/400 lineup, the 9404-B20 deskside system, the closest analog to today’s Power 720 machine, which is also the biggest seller for supporting the IBM i operating system.

The B20 machine could have from 4 MB to 16 MB of main memory and has a proprietary CMOS processor rated at 5.1 CPWs on the Commercial Processing Workload benchmark, which wouldn’t even be created until 1996 and which is still used today to gauge relative performance. The system had two 315 MB disk drives, and if you wanted to splurge, you put in a single 945 MB disk. The B20 had a single I/O bus and four peripheral slots for plugging in twinax adapters, network cards, modems, and other devices. The B20 cost $33,500 in a barebones configuration, and an OS/400 license cost $10,500. If you bought a preconfigured machine known as a Total System Package, IBM cut $2,400 off the base system price. A 4 MB memory stick ran you $6,000, and a 315 MB disk drive cost $5,500.

Now let’s look at the Power 720, and we’ll take a base machine of the Gen2 vintage (with PCI-Express 2.0 peripheral slots, announced last October) with four 3 GHz cores and we’ll only put IBM i 7.1 on one of the cores. This machine offers from 4 GB to 128 GB of main memory, and doesn’t come with disks in its base price but it has room for eight drives and a total of 4.8 TB of internal disk capacity in its six disk bays. There are five PCI-Express 2.0 slots plus another four optional low-profile slots if you use a riser card, and two 12X I/O slots if you want to hook out to external disk. So basically, you can make this Power 720 look like a big bad box with external disks, something like a B60 back in 1988 or a B70 back in 1989.

But let’s keep it simple and put in 8 GB of memory and two 139 GB disks while activating all four cores on the box in an Express Edition configuration. That would cost $6,409. An IBM i 7.1 license on one core costs $2,245 for five users, and then you need to pay another $2,500 to add ten users for a total of 15 users on the system–about what you’d expect to be on a B20 machine sitting in an office back in 1988. Add in another $750 for a full year of software support, which used to be part of the OS/400 license. (The assumption was that you would upgrade to new versions and pay for software upgrades and that you paid monthly hardware maintenance. All the numbers worked out in the end.) Add it all up, and you are talking about $11,904 for a machine that delivers approximately 6,000 CPWs. (IBM’s performance metrics are 23,800 CPWs for a four-core machine, and it doesn’t give ratings for a single-core setup.) If you want to be even more fair, you could say that customers back in 1988 would probably have the machine for at least three years before upgrading, so you should add three years of Software Maintenance to the Power 720 to make it more apples-to-apples on the comparison. That puts you at $13,944. An 8 GB memory stick runs you $1,065 on the Power 720 and a 600 GB SAS 10K RPM disk costs $1,200.

That Power 720, with just one core activated, has nearly 1,200 times the aggregate transaction processing throughput of the B20, and its capacities on every scale are much larger. Maximum compute capacity is actually 4,700 times greater when you activate all four cores and run IBM i in a single system image across it. That is the impressive leap that is made possible by the combination of intense competition over 24 years, with not much room to stop and breathe, and the thus-far relentless advance of Moore’s Law on chip manufacturing processes. Maximum main memory on the Power 720 is 8,192 times greater than on the B20, and disk capacity is a little over 5,000 times as great. That B20 cost $41,600 and delivered a unit of CPW performance for $8,157; the Power 720 does it for $2.33, or a factor of 3,500 times better bang for the buck. If you want to do a four-core Power 720 in a base configuration, with all of the cores running IBM i, then the base machine would cost you $29,554 to deliver 23,800 CPWs, or about $1.24 per CPW. If you loaded up the Power 720 with memory and disk sufficient to actually drive all four of those cores doing transaction processing work, it would still be less than $2 per CPW.

Clearly, the capacities have grown faster than the prices have come down, but the prices have actually come down. To be fair, Java is a pig and so is PHP–at least by RPG and COBOL language and 5250 protocol standards–so it takes a lot more CPU, memory, and I/O capacity to get the same transaction processed in real-time or the same invoice generated in batch mode. But still, the advance of the technology is no less remarkable because of that, and this is the price you pay for the ability to write more flexible RPG and COBOL code or to be portable with Java or PHP code. It is also the price you pay for a system that can run IBM i as well as AIX and Linux natively, and that can sport PowerVM hypervisors to virtualizes machines and drive up utilization.

That the AS/400 exists at all, by whatever name and in whatever form, after 24 years in this cut-throat, take-no-prisoners IT business is nothing short of remarkable. Having seen a bit of human nature and studied some history, I am not about to make any silly predictions that the Power Systems-IBM i combo will be around for another 24. None of us can know that–not even IBM. We only sort this out together, each day, with the choices we make about how we do our computing. Or don’t.

But I will say this. There are still OpenVMS and MPE machines out there, just like there are still Unisys and Fujitsu mainframes, or even System/3X machines and an even larger number of AS/400 and iSeries boxes. How many people predicted the AS/400 and the mainframe would be dead, both economically and technically, by now?

As far as I can tell, the IBM i platform is paying for itself and IBM seems content to let its customers keep running their RPG and COBOL code on Power iron and is happy to make tweaks and do tunings to the IBM i platform to support new iron and new software technologies. The company included IBM i in the PureSystems and Flex System hardware announcements, although it needs to get its SmartCloud public act together with IBM i.

The important thing is that Big Blue will support IBM i as long as it can make money at it–and perhaps even a little longer. And that is about as good as it ever gets in the systems business. So let’s get on with year 24 and get some work done.