NASA PC cluster on IT crime trail

NASA has designed a low-cost, high-performance computing cluster of PCs to analyze evidence gathered from network intrusions and other computer crime investigations.

The space agency's Computer Crimes Division (CCD) is using a Beowulf-class high-performance computing cluster to process evidence gathered from computer-crime victims' machines and computers confiscated from suspected computer criminals, said Thomas Talleur, the advanced technology programs executive.

The Beowulf project, pioneered at NASA, studies the advantages of harnessing the computing power of interconnected PCs to process the kind of large volumes of data traditionally handled by expensive supercomputers.

The system's ability to process massive amounts of data will be used to speed the work of computer-crime officials when analyzing the hundreds of gigabytes of disk space the division often acquires from a suspected criminal's machine, Talleur said.

"This is a revolutionary leap in performance," he said. "It's going to cut down months of work into an hour or less. We're looking for text— what the bad guy's footprints in the sand are. We put large images and large data sets up and search it all at once in the kernel of the operating system."

For example, the CCD recently seized hundreds of gigabytes of data and images from an Israeli hacker suspected of hacking into NASA systems.

It took officials seven weeks to cull through the data using standard computers. With the new system, the same work could be done in seven minutes, Talleur said.

Operate as Single System

NASA engineers adapted the powerful, Unix-like operating system known as Linux to work in a massively parallel and distributed computing environment using commercial off-the-shelf hardware. The idea is to allow the multiple PCs to operate in concert as a single system.

Linux is distributed for free with a wide variety of robust tools, allowing users to adapt or create powerful, customized computing solutions. The CCD cluster was built for less than $60,000. More traditional supercomputers that use raw computational power, as opposed to linked processors, can cost hundreds of millions of dollars.

While NASA has used Beowulf-class computers in various scientific applications, this is the first time the high-performance alternative has been used for non-scientific applications, Talleur said.

Debra Goldfarb, International Data Corp.'s vice president of workstations and high performance systems for IDC Research, said there are only a few applications that can be strategically aligned to work on Beowulf-type parallel systems. Applications well-suited for these types of systems include image processing, query search and rendering, which is most commonly used in the entertainment industry.