IBM Fired Up About Power7-Based Smarter Systems

Published: February 15, 2010

by Timothy Prickett Morgan

You are probably wondering why IBM started the Power7 systems rollout in the middle of the line instead of rolling the line out at once as it did in days gone by with the AS/400 and RS/6000 machinery. It is really simple. IBM has dominant market share in the high-end of the midrange lineup represented for the past six years by the 570-class machines--iSeries, pSeries, System i, System p, and Power Systems all--and the company wants to keep it that way.

According to Ross Mauri, general manager of the Power Systems division within IBM's Systems and Technology Group, the 570-class machines have over 65 percent market share. I was taken aback a bit by this revelation.

Mauri did not elaborate on precisely what market the share was coming from, but this being Power Systems, he no doubt meant percent of revenues for Unix-based machines in this high-end midrange price band. Perhaps with some i/OS system sales thrown in for good measure because IBM is starting to talk like i/OS is, being a very reliable operating system that has POSIX compliance and supports almost all of the SPEC 1170 features to qualify as being called a variant of Unix (as IBM's z/OS can be called, too, by the way, because it does support all the SPEC 1170 features). Maybe we can just rename it AIX/400 and get it over with? Perhaps call ILE RPG something clever like Db to contrast it with Microsoft's C# Java-alike and let Db run inside its own virtual machine, or even the Java virtual machine like Groovy and some other languages can now do.

But I digress. Frequently, and with enthusiasm.

Like all chip and system makers--and IBM and Oracle are the last two big ones who do both, including operating systems, databases, and middleware--Big Blue has been scratching its head for the past decade, looking for a way to keep the transaction workloads growing in the world. This is important because Moore's Law keeps pushing down the price of a unit of processing and storage at an alarming rate, and without an equally large increase in the complexity and volume of workloads, no one can make any money in this IT racket. When the profits run out, so do the technological advances. In this chicken and egg theory, the egg came first, and in this case, it was laid by a dinosaur with pretensions of color and flight.

After wandering around in the marketing desert like the original evangelist, muttering about eBusiness, On Demand, Dynamic Infrastructure, and a bunch of other buzzwords I have thankfully forgotten, IBM has come to the latest--Smarter Planet--as the salvation for it woes, er, excuse me, I mean our woes. The Smarter Planet marketing campaign is real, in the sense that it is meant to move politicians to back massive projects to automate infrastructure such as roads, railroads, shipping, electrical power, water supply, communications, and maybe even healthcare and politics if you count those as basic infrastructure. (I do.) Smarter Planet reminds me of Kentucky Senator Henry Clay's American System and its constant banging on the drum for "internal improvements" funded by the Federal government through protective tariffs. Incidentally, Smarter Planet, which is interested in upgrading and automating basic infrastructure in countries around the world, is most definitely not a counterweight to the laissez faire capitalism that ruled in Britain following the War of 1812 and up through the Civil War in the States, when the American System was thought of as a means to purposefully bind the country together through infrastructure. Better transportation and eventually communication networks would help everyone lead better lives.

When IBM is talking about Smarter Planet, the talk is about improving the world, to be sure, but it most certainly is not meant to be a counter balance to free trade and globalization, the modern instantiation of laissez faire economics. Smarter planet is about selling servers, software, and services.

To do that, IBM and its rivals are going to have to convince the governments and corporations of the world to gather lots more information in the hopes of running infrastructure more efficiently, thereby saving money. The Power7-based Smarter Systems announced last week in New York are, says IBM, the culmination of three and a half years of work and are just the kinds of machines that will be needed to process the massive amounts of telemetry gathered up from trillions of network-enabled devices that are part of this Smarter Planet IBM wants us to live in.

"The world is becoming more instrumented and more interconnected," explained Rod Adkins, the senior vice president and general manager of IBM's Systems and Technology Group, at the launch event last week. "And that is going to affect every system in the world."

"This is not a chip announcement, it is a systems announcement" Adkins continued. "This is not a reaction. This is not a new strategy based on a recent acquisition," Adkins said referring to Oracle's finally consummated acquisition of Sun Microsystems at the end of January. "This is not a bag-of-parts announcement," said Adkins, taking two jabs at new system-ish competitors: one at the Acadia partnership put together by Cisco Systems, VMware, and EMC to sell virtualized server and storage stacks called Vblocks, and the other at Hewlett-Packard and Microsoft, which are assembling Windows-based x64 stacks called Frontline.

Here is the kind of workload growth that has Big Blue salivating. IBM is partnering with eMeter, which makes software for culling data from smart meters behind electric, gas, and water utilities and putting it to immediate use for managing those grids as well as for long-term use to send you a bill. The company's software is currently used on over 24 million meters, but there are billions to go in this Smarter Planet scenario.

One of eMeter's customers is a power company in Ontario, Canada, that is looking to put smart meters out there for 10 million customers. Using manual meter reading from analog meters--meaning a human being walks around every month or two to write down the numbers on the spinning meter so they can be entered into the billing system--means processing 120 million transactions per year for monthly data, which weighs in at around 30 terabytes for the whole year. That's not much of a big deal for a modern system, really. Moving to smart meters so electric use can be tracked on a daily basis means processing 3.65 billion transactions per year for those 10 million customers, and generating nearly 1 petabyte of data, according to Adkins. The goal at this electricity distributor, however, is to read the smart meters every fifteen minutes, not once a day, and that means needing to plow through a little over 350 billion transactions per year and something on the order of just under 100 petabytes of data.

"There's a new type of performance that is required here," Adkins said, adding that none of this includes the data analytics that would be required to use this polling data from the smart meters.

Let's do some math for fun. Consider that maybe somewhere around 2 billion households worldwide could be wired up with smart meters. Then you're talking about a factor of 200,000 or so more transactions and data for putting smart meters on all electric grids compared to the 10 million outlined above. We're talking about 20,000 exabytes of data per year, and you're talking about just under 70.1 quadrillion transactions to poll those meters. That works out to 133.2 billion transactions per minute for the entire globe, in aggregate, for every minute of the year, sustained. IBM's current Power 595 boxes can handle on the order of tens of millions of relatively simple transactions per minute, so such efforts will require lots of servers, even with Power7 and Power8 machinery. Just to support this workloads would take, by my estimate, somewhere on the order of 3,000 of the top-end Power 795 servers IBM will deliver this year with 256 cores, at a price tag of around $4 billion to $5 billion, including storage and assuming a fair amount of flash memory in there, as well as heavy discounting.

To put that into perspective for you, that is an entire year's worth of Power Systems sales, just for smart meters, and not including processing capacity for analytics. That is just to poll the 2 billion meters worldwide.

You can see now why IBM is so excited.

Not being an application provider any more, as it was in the salad says of the mainframe and midrange market, IBM is going to have to partner tightly with ISVs like eMeter to get those smart infrastructure deals and to keep other companies--namely Oracle as well as the Acadia and Frontline partnerships and whoever Dell gangs up with--from stealing its share of the business. (Which I reckon is somewhere on the order of 25 percent or so, based on IBM's rough share in servers and storage.) So what IBM needs to do--and what AS/400 shops have long understood--is tune an entire stack of hardware, systems software, and application software so it can win these infrastructure instrumentation and analytics deals.

To that end, Adkins says that IBM rejiggered its sales force in 2009 in preparation for the 2010 Power7 systems rollout, and sales reps are being compensated more for solution-based selling, not just pushing raw iron. "You will see us do a lot more active partnering and selling with our ISVs," Adkins said.

I would not be surprised to see IBM bite the bullet and merge with SAP, and thus go directly at Oracle. But such a merger would indeed be that: a merger, not an acquisition. Even after taking hits in recent months, SAP still has a $51.5 billion market capitalization, compared to IBM's $160.9 billion as I write this. IBM could stop buying back its shares for a decade to pay for the deal, of course, and do an outright acquisition of SAP. But who is gonna loan Big Blue $70 billion in the meantime to do the deal? Uncle Sam? Goldman Sachs? (Is there a difference?)

An IBM-SAP combination would give IBM three seats on the board for every one of SAP's, so Big Blue would still be in control. The IBM-SAP combo would have had $110.4 billion in sales in 2009 (assuming current dollar-euro exchange rates) and just over $17 billion in net earnings; SAP would make up about 13 percent of revenues of the combined company and about 21 percent of profits. Such a deal would be immediately accretive to IBM's profits and would, I think, cause Oracle and the entire industry to take a deep breath. And antitrust regulators and the German government in particular. IBM might even sponsor an America's Cup boat just to be extra mean.

Smarter Systems is going to require a smarter strategy. And that means admitting that Oracle was just as right about controlling applications as it appears to be about wanting to control its own hardware and systems software stack.

How the AS/400's progeny would not get lost in such a monstrous reorganization of IBM is beyond me. Those i/OS shops will just keep a-doing what they have always been a-doing, which is coding their own applications and doing their own thing, I suppose. As long as IBM supports i/OS on future Power Systems iron, and keeps investing in development for the database and other features, this is what really matters to most i/OS shops.

RELATED STORIES

The Power7 Rollout Begins In The Middle

The i/OS Roadmap Revealed--Sort Of

http://www.itjungle.com/tfh/tfh020810-story03.html">The System iWant, 2010 Edition: Blade and Cookie Sheet Boxes

The System iWant, 2010 Edition: Entry Boxes

The System iWant, 2010 Edition: Midrange Boxes

IBM Preps Power7 Launch For February

Looks Like i 7.1 Is Coming In April\r\n \r\nThe System iWant, 2010 Edition: Big Boxes

Power Systems i: The Word From On High

Power Systems i: The Windows Conundrum

Power Systems i: Serve's Up

Power Systems i: Thinking Inside the Box

Rolling Thunder Rollout for Power7 Processors Next Year

IBM Rolls Up an i 6.1.1 Dot Release

The Curtain Rises a Bit on the Next i OS, Due in 2010

Start Planning for Power7 Iron Now

IBM to Reveal Power7 Secrets at Hot Chips

Power 7: Lots of Cores, Lots of Threads

                     Post this story to del.icio.us
               Post this story to Digg
    Post this story to Slashdot