Big Blue says Power7 will make world smarter
It's not a chip. It's a system
If you believe that putting sensors on all manner of infrastructure systems, gathering telemetry on them in real time, and automating how they perform to the nth degree will all make the world a better place, you're well suited to a job at IBM selling its vision of a Smarter Planet - and some Power7 servers.
The top brass of Big Blue's Systems and Technology Group pitched the Smarter Planet vision yesterday at the company's Power7 launch in New York. The basic idea is that key infrastructure systems - transportation, electricity, and water - will need massively more powerful iron to gather up telemetry and use it to run this infrastructure more efficiently and that IBM is the vendor you could trust to deliver that big iron. IBM, you see, sells systems, not just commodity servers.
Of course, if IBM is wrong, then it went to a lot of trouble to create more powerful boxes that will do little more than consolidate footprints among its AIX customer base and entice some HP-UX and Solaris shops over to the AIX fold over the long haul. (This footprint contraction has most certainly happened in its mainframe and AS/400 customer bases, by the way).
But no one wants to talk about how elastic (or inelastic) demand for processing capacity will be in the Unix market, particularly once new iron from IBM, Hewlett-Packard, and Oracle starts grinding against each other around the middle of the year like the action on the dance floor in a private club in the Bronx. (Wait, that's the wrong metaphor, but you know what I mean). All that IBM knows for sure is that the Unix market is somewhere around $14bn to $15bn and that at $4bn each, Sun and HP have money that IBM wants.
"This is not a chip announcement, it is a systems announcement," said Rod Adkins, senior vice president and general manager of STG. "This is not a reaction, this is not a new strategy based on a recent acquisition," Adkins said referring to Oracle's $7.4bn consumption of Sun, turning the company into a hardware maker and one that surprisingly says it is committed to building application-tuned systems.
"This is not a bag-of-parts announcement," Adkins continued, referring to the Acadia partnership between Cisco Systems, EMC, and its VMware minion as well as the Frontline partnership between HP and Microsoft.
IBM throws statistics around like Sun and Intel when they talk about pervasive computing, and Adkins fired off some numbers in his presentation: over 1 trillion connected devices on networks by 2011, over 2 billion people on the Internet, and a 10X growth in data, most of it unstructured. "The world is becoming more instrumented and more interconnected," explained Adkins. "And that is going to affect every system in the world."
To give a sense of the kinds of loads that modern systems are going to have to do in this massively instrumented world, Adkins trotted out some statistics from eMeter, a partner of Big Blue's that is involved in smart electrical grid projects. A utility company in Ontario is looking at installing 10 million smart meters.
Currently, the systems behind the electric power distribution company (which was not named) processes 120 million transactions per year taking in data gathered in from analog meters once or twice per month. (And using those pesky humans with their lunch breaks and health insurance to gather the data).
Installing smart meters at this company and polling the meters once a day - giving the power company fine-grained data and therefore fine-grained control over power generation and distribution - will drive up the transaction loads on the same applications used for calculating the electric bill to 3.65 billion transactions per year and generate over 1 petabyte of data. And polling those smart meters every 15 minutes, as the utility company things is necessary to really do smart grids, will push up transaction loads for keeping track of electric use and generating the bills will skyrocket to 350 billion transactions per year.
IBM is looking at demonstrating that it can build a system that can handle the smart meter loads for 50 million smart meters, which is ten times this load.
Sponsored: Hyper-scale data management