The Friedrich Miescher Institute (FMI) tamed its rapidly growing research data and solved its backup problem by...
By submitting your email address, you agree to receive emails regarding relevant topic offers from TechTarget and its partners. You can withdraw your consent at any time. Contact TechTarget at 275 Grove Street, Newton, MA.
combining a Copan Systems virtual tape library (VTL) and Sun Microsystems' SAM-FS Storage Archive Manager to complete a "paradigm shift" away from "dead-end traditional storage filers."
FMI, based in Basel, Switzerland, specializes in research for the molecular biology of disease. The institute captures a massive number of large files. Most of the data becomes persistent within 30 days, but occasionally files have to be accessed quickly for analysis.
With storage requirements projected to grow from 10 TB in 2006 to around 800 TB by 2016, FMI formed a committee to come up with the best way to handle the needs of its 300-plus users in 22 research groups.
"A couple of years ago we had quite a storage challenge," Flanders said. "It became clear this data is always going to increase and increase and increase at a rate we can't control or anticipate. I had a NetApp order in. I said, 'If we go with this it will work and nobody will get fired for working with NetApp.' The committee said, 'Are you sure this will work long term?' I threw the order away."
Flanders wasn't just soured on NetApp. He decided no traditional NAS filer could meet FMI's needs. FMI selected an EqualLogic (now part of Dell) iSCSI SAN for transactional data, and the Copan VTL and Sun SAM-FS combo for files.
"I said we need to approach this differently. We needed a paradigm shift. Traditional file servers don't scale, they're just a dead end. So we married SAM-FS and Copan [VTL] and created an infinitely scalable file system," Flanders said.
Flanders bought a Copan Revolution 300T VTL late last year, mainly because of its density and MAID spin-down. FMI started with 40 TB, and Flanders said he's looking at a 40 TB or 80 TB expansion.
"We looked at traditional file server approaches, and we also looked at EMC's Centera. That's a cool solution, but it's a dead end, especially with our capacity," Flanders said. "We have projections into the petabyte range."
At the suggestion of the San Diego Supercomputer Center, he went with Sun's SAM-FS. SAM-FS copies files to several targets to archive and backup data simultaneously. Administrators set a quota, and when a volume hits the quote, it releases data from the file system to free space. When a user accesses data, SAM-FS copies it back to the cache and streams it from the archive or backup. That prevents volume sizes from growing too large.
"I used to divide LUNs for different research groups, so each group had its own file system," Flanders said. "But they became large and unmanageable. With SAM-FS, we have one 40 terabyte file system. We make a 3 TB cache on Copan, and I give each research group a 1 TB or 2 TB quota."
Flanders said using SAM-FS to write data to Copan's VTL and SpectraLogic T950 tape drives alleviates the need for daily backups.
"We eliminated backups," he said. "SAM-FS automatically backs itself up. When people write data into the cache, a copy is taken into the VTL and another copy goes into tape. Each piece of data is written into the cache to VTL every hour and into tape about every 12 hours. For disaster recovery, we have a VTL and two tape libraries."
The decision to go with Copan over more established vendors was made easier because Copan uses FalconStor software for its VTL, Flanders said. That's the same software EMC, IBM, Sun and other large vendors used at the time.
"Everybody and their brother uses FalconStor," he said. "It was a known quantity. If Copan had written its own software, I probably wouldn't have bought it."
Flanders said information is available on the Copan VTL anytime. When a user tries to access a file on a Copan drive that isn't spinning at the time, there is about a 20 second delay. However, "Most of these apps can handle a 20 second delay," Flanders said.
The most obvious advantage for Copan is power and cooling savings from MAID. "I can fully load that thing, and I have 25 kW to 30 kW in my server and have almost 800 TB," he said. "We have 110 [SATA] disks in the tray, and in a year we've had one disk failure."