MIT Puts Storage Under Microscope

Talk about a granular line of business.

So as the Whitehead Institute-MIT BioImaging Center at the Massachusetts Institute of Technology (MIT) creates image archives of individual molecules, it struggles to keep up with storage requirements.

The institute recently expanded its Archivas cluster system from 10 TByte/s to 50 Tbyte/s after several months. (See Archivas and Archivas Seeks Archiving Action). Next up, it will look to upgrade the Silicon Graphics Inc. (SGI) SAN it uses for primary storage.

According to its Website, the Whitehead Institute-MIT BioImaging Center unites the power of leading-edge microscopy and advanced computational systems to study the structure, dynamics, and function of molecules in challenging problems faced by biology, medicine, and bioengineering.”

Or as research scientist James Evans puts it, “Our research is a basic understanding of biology -- how the human body works. People understand how genes work, but not how they fight disease or go wrong when cancer is formed. That’s what we try to understand. Imaging is a good source for understanding systems. Say you wanted to understand how a car works, if you took a movie, you’d see how the wheels turn.”Whether it’s in Hollywood or in a Cambridge, Mass., lab, making movies is expensive. Whitehead uses about 10 different million-dollar microscopes to take snapshot images of molecules. The resulting digital images fill up disk drives fast.

“There’s a lot of data generated from imaging,” Evans says. “We generate terabytes of data in a month. We have instruments that can generate 1 Tbyte in a day. And we have 96 faculty members generating a little data here, a little data there. There’s a huge demand for storage.”

Evans, who helps evaluate storage technology at Whitehead, says his group has about 15 Tbytes of primary storage on a SGI Fibre Channel SAN. It not only maxes out fast, but then all the research data must be retained so it can be reviewed after it’s published. To archive data, Whitehead implemented a system with Archivas Cluster (ArC) software and SATA drives last summer.

“With Archivas we can move data off our high performance storage, free up space, and get it back with reasonable performance,” he says.

Although SGI isn’t among the storage leaders, it does have a foothold in high performance computing (HPC). (See SGI Wins at SC|05.) But Archivas is a newcomer in storage, and competes against the industry’s leading players such EMC, Heweltt-Packard, IBM, Network Appliance, and Sun, as well as startups Nexsan and Permabit.Evans says he looked at EMC’s Centera as well as “all the other” major players in a two-year evaluation process before picking Archivas.

“Being an academic institute, we need something low-cost, easy to manage, and scalable,” he says. Archivas permitted the organization to start with 10 Tbytes, without paying hundreds of thousand dollars for something more. "We wanted to be able to start small and grow really big," Evans adds.

Archivas software lists for about $75,000 for 10 Tbytes of managed storage.

Unlike Centera and other CAS systems, Archivas sells software that runs on off-the-shelf hardware. When it came time to upgrade, Whitehead added licenses and installed more SATA drives. “We’re not locked into any company’s brand of controllers, so we don’t have to wait for them to update those,” Evans says.

The only problem is getting Archivas to work flawlessly with some of Whitehead’s custom-built instruments. “We’re working with them to get better performance with our automated microscopes,” Evans says. “It takes time. We have instruments a little different than what their other customers have.”— Dave Raffo, Senior Editor, Byte and Switch

Organizations mentioned in this article: