Medical Archive Grows Its Own Grid

Standards are great and blade servers have their place, but the U.S National Digital Medical Archive (NDMA) rejected both and instead built a grid for millions of hospital records using traditional hardware and a proprietary grid architecture.

The commercial archive, operated by Berwyn, Penn.-based i3 archive, stores around 1.5 million images for 30 hospitals around the U.S, including mammograms, CAT scans, and MRIs.

Derek Danois, the firms president, said that despite the hype surrounding blade servers, he has no plans to move away from the standard IBM xSeries servers currently in use. “Our processes, believe it or not, are not computationally intensive,” he says, adding that only 20 percent of the machines’ capacity is currently used.

The archive is powered by a 64-node grid of xSeries machines, hosted by IBM in two New Jersey and California data centers. “Their primary function is load balancing and traffic cop duty for data as it goes over the grid,” says Danois.

Despite a rash of recent announcements from vendors, users are sending a mixed message about blade technology. (See IBM Flashes New Blades and Blade.org Adds Voltaire .) On the one hand, some organizations are wowed by the processing power and space savings offered by blades, whereas others have balked at cost and heat issues. (See NewEnergy Chops Its Blades, Study Highlights Blade Disappointment, and Blades for Buffalo .)In addition, i3 also opted against the Globus toolkit, an open-source middleware toolkit for building grid systems and applications. (See Vendors Form Globus Consortium and Keynote: Grids to Grow.) Danois says that back in 2003, when i3 began work on the archive, this made perfect sense. “At that point [the toolkit] only focused on computational transactions, and what we were facing was a massive I/O challenge -- distribution and storage of large packet data."

i3 got round this problem by developing its own set of standards. “There was a different architecture that had to be put forth here -- a proprietary database schema using [IBM’s] DB2 database,” says Danois. While this lets i3 manage the archive’s data repository, he adds, the firm also developed APIs to enable communication with other grids.

Danois is not the first person to highlight grid standards issues. Other users have bemoaned a lack of standards in the grid space; some say it's hampered the technology’s progress. (See Grid Computing: Baby Steps.)

With the grid taken care of, however, i3 is now planning to expand the archive’s storage to support an estimated 10 million images by the end of the year. Some 50 Tbytes of images are currently stored, consuming about half the current 100-Tbyte capacity. “By the third quarter we will be using as much as 300 to 500 Tbytes,” says Danois

IBM execs must be rubbing their hands with glee. “There’s potential for us to be procuring $50 million in year in storage services from IBM within three years,” explains Danois, a massive increase on the $3 million i3 spent with the hardware giant last year.But Danois admits that there are pluses and minuses to using a hosting partner. “It’s simply easier to let them manage -- they can handle things like firewall matters and upscaling bandwidth. The challenges are some of the response times. Sometimes it’s half a day, and, as a growing company, we need a faster response than that.”

Nonetheless, Danois is looking to extend the archive’s reach beyond healthcare. ”I think that the technology is applicable to other large volume datasets. We’re exploring some opportunities, particularly in media and multimedia."

i3 told Byte and Switch that IBM was the logical choice to host the archive, explaining that the hardware giant had been involved in the project since the late 90s, when the National Institutes for Health first unveiled the scheme.

— James Rogers, Senior Editor, Byte and Switch

Organizations mentioned in this article: