IBM: Blue Gene Saves Energy, Space

NEW YORK—Customers across the board are looking for systems that consume a lot less power and take up less floor space, which is creating a huge opportunity for IBM to promote its Blue gene supercomputer to a range of industries, including the financial sector on Wall Street.

In an address entitled "IBM Blue Gene: ultra-scaling and power efficiency for computational advantage in financial services," at the Linux on Wall Street conference here April 24, Dave Turek, the vice president of deep computing at IBM, said energy, floor space and the ability of systems to be recycled are the biggest issues facing customers today.

Click here to read more on how AMD and others are looking to address server power consumption.

Customers all also want real-time solutions to their problems and issues, but thats where the commonality between them ends, he said, pointing to the fact that theres no such thing as one size fits all. "There is such a thing as one size kills all, as that produces something customers no longer want," he said.

Resource Library:

Open source and Linux are catalysts to innovation and have been embraced by IBM. "It is important to embrace the future as if you do not you can end up in a dead end alley," he said, adding that there are more things coming from IBM as it moves toward petaflop systems that can handle trillions of calculations a second.

IBM System Blue Gene was first announced in 1999 and is the worlds fastest supercomputer. It scales up and down, producing sustained performance in applications in excess of 100 teraflops, Alan King, a research member at IBM and a user of Blue Gene, told the approximately 100 attendees.

"IBMs system agenda is one that is based on collaborative innovation, openness, virtualization and technology innovation that matters. Openness is the way we leverage our technical assets in the marketplace," he said.

A Blue Gene system starts off as a two-system processor built onto a card. These are then loaded onto racks and one rack gets 2,000 processors and can run as if there were 1,000 nodes, he said.

"The way Blue Gene works its magic is the multiplicity of networks built into the back end. With 65,000 processors, an intelligent network was needed to work around system failures, so there was a lot of effort directed toward making this a reliable system as well as a fast and efficient one," King said.

Blue Gene brings 1,000 nodes or 2,000 processors together in the space of a single rack. Each node comes with a gigabyte of memory, translating into a terabyte for a single rack, which costs between $1.5 million and $2.0 million. "As such, these systems bring an enormous amount of capability," he said.

Click here to read more about IBMs Project Fastball.

Blue Gene represents an innovative way to scale multiteraflops of capability, bringing leadership performance and price/performance, massive scalability and efficient packaging that drives low power, cooling and floor space requirements, King said.

"Blue Gene is also unique in the market for its balance between massive scale-out capacity and preservation of familiar programming environments. It is also applicable to a wide range of computationally intensive workloads such as financial instruments pricing and risk analytics," he said.

But IBM is hearing from customers that Blue Gene needs to integrate with grid middleware, so the company will be making some announcements with grid middleware vendors in this regard in the next two months, he said.

While Blue Gene runs a stripped-down version of Linux that can handle file descriptors and multiple threads, establishing relationships with grid middleware vendors will allow users like those in the financial sector to just slot it into their environment by allowing them to do the data services management as well, King said.

Editors Note: This story was updated to correct the details of the systems pricing and number of processors.

Check out eWEEK.coms for the latest open-source news, reviews and analysis.