IBM Sees Green in Going Green in Data Centers
May 14, 2007 Timothy Prickett Morgan
In New York City last Thursday, when the weather was just starting to get a little muggy but the air conditioning was cranking so high in the Helen Mills Theater that you could just about hang sides of beef on the stage, IBM announced a new initiative, dubbed Project Big Green, under which it will sell technology and services to help the world’s data centers consume less power as they support expanding IT workloads.
While IBM as well as other server and storage array makers have been conscious of power and cooling issues surrounding IT gear since there were data centers, the issue has come to a head in recent years as dense X64 servers and myriad pieces of networking equipment have stormed the data centers of the world and taken over. When blade servers entered the market seven years ago, they were the straw that broke the air conditioner’s back, since the power density of blade servers means companies can pack a lot more computing into an area., generating heat islands that are tougher to deal with compared to more spread out computing architectures of the past.
Under the Project Big Green initiative, IBM is reallocating $1 billion in resources previously earmarked for other developments toward making data centers more energy efficient. The company has organized more than 850 employees with various levels of expertise in IT architecture and data center design and trained them up as energy efficiency architects who will work in conjunction with its Global Services arm to sell services to help data centers cope with heating and cooling issues. Obviously, IBM’s System and Technology Group has a big hand in Project Big Green, given that it designs and markets its servers and storage as well as the systems software that runs atop them and, in some cases, such as with its PowerExecutive tool, actually puts a governor on energy usage in IBM servers.
Project Big Green is being spearheaded by Val Rahmani, general manager of Infrastructure Management Services within the Global Technology Services half of the Global Services giant that makes up more than half of IBM’s sales each quarter. Rahmani was the assistant to former chairman and chief executive officer Louis Gerstner, and ran IBM’s Unix server business and its sales and distribution organization prior to this assignment at Global Services. Her boss is Michael Daniels, senior vice president of Global Technology Services, who is in charge of IBM’s own data centers as well as those that Big Blue runs on behalf of its outsourcing clients; the GTS half of Global Services also supports SMB shops and collects IBM’s maintenance money. Bill Zeitler, senior vice president in charge of Systems and Technology Group, was with Rahmani and Daniels at the Project Big Green launch last week, and they went over a lot of statistics about how bad the energy issues are at data centers today.
“Climate issues are top of mind today,” said Rahmani. “The energy crisis is not something abstract to us in the IT industry. This is real, and it is happening now.” She cited figures from IDC that claim for every $1 that companies pay for computer hardware, they pay roughly 50 cents for the energy to power the machinery and to cool it. In the next four years, IDC estimates that energy bill is expected to increase by 54 percent to 71 cents for energy for every $1 spent on hardware. “It is now coming to the point where no one can ignore the issue any more. And we’re mobilizing IBM,” Rahmani said.
IBM has some experience in managing data centers, with 30 million square feet of current data center space that it has “influenced,” according to Daniels, and 8 million square feet of its own over at Global Services. Like other outsourcing companies in particular and a growing number of corporations with large data centers in general, Global Services wants to get more work out of existing machinery, and because of the explosion in applications, it is also projecting that it will need to add more servers and storage, too. So IBM is turning Project Big Green on itself first, and has committed to doubling the processing capacity of its data centers in that 8 million square feet space–and doing so with half the number of computers. The end result will be that IBM does not have to build an additional 8 million square feet of data center space, which requires a huge outlay in capital for real estate and construction as well as a doubling of the electric bill.
Most data centers are dealing with an energy issue that is on a more manageable scale, but the problem is no less severe, according to Daniels. He says that the average 25,000 square foot data center spends about $2.6 million a year on power, and that over the long haul, IBM reckons it can help companies cut about half of their power usage without sacrificing performance or capacity.
To accomplish this requires a variety of technologies, including using blade servers where possible because of their integrated cabling and switches; moving to fiber optic cabling under the raised floor of the data center, which keeps air flow moving; using water jackets like the Rear Door Heat eXchanger, which grab heat close to servers and moves it to chillers rather than letting it dissipate into the data center; and installing other new technologies such as a chiller, called the Data Center Stored Cooling Solution, that IBM created for its own data center outside of Bromont, Quebec. This latter item basically makes ice at night when the electric usage is low and the price is lower and uses that ice to cool the data center during the day, when the applications are burning more juice and electricity prices are higher. IBM has other water-cooling technologies coming to market later this year, too. IBM is also advocating the use of specialized server appliances, such as the DataPower XML accelerators it sells, or the Cell chip, for number-crunching, for particular workloads so the energy per unit of work goes way down compared to general purpose servers.
And finally, of course, server virtualization is the low hanging fruit, with X86 server utilization at an average of around 8 percent and Unix server utilization at maybe 20 percent, compared to 80 percent for mainframes. “We want to get more work out of each individual unit of capacity you install,” said Zeitler.
As an example, Zeitler talked a bit about Pacific Gas and Electric, the San Francisco-area power company that supplies juice to about 40 percent of California’s population and a dense area of businesses. PG&E is paid not on how much energy customers use but on how efficient its energy operations are, because California was hit with an Enron-induced energy crunch seven years ago. PG&E, which is the first power company to give rebates to corporate customers that deploy energy-efficient servers, also has a 40,000 square foot data center that uses a lot of power, and because of this, it is consolidating some 300 Unix servers down to six of IBM’s top-end System p5 595 Unix boxes, and thereby reducing the power consumed by these servers by 80 percent while still supporting the same applications, but on logical servers instead of physical servers.
Of course, going green in the data center will require the services expertise of Big Blue–just like any technology seems to these days. (Funny how that works, eh?) IBM is selling data center assessment services to help companies figure out how much power they are using, where it is going, and how they can reduce it. Heaven only knows what these services cost, because IBM does not provide list prices for services. IBM is also selling what it calls the Scalable Modular Data Center, which is a 500 or 1,000 square foot data center setup, complete with servers and storage, which can be plunked down in eight to 12 weeks and which is about 15 percent less expensive than piecemeal building out of the data center.