NCSA planning 500 petabytes of storage capacity

05.22.12 -

The National Center for Supercomputing Applications (NCSA) is adding 380 petabytes of automated, near-line storage capacity for data produced by scientists and engineers using the sustained petascale Blue Waters supercomputer and to support other projects. The center plans to eventually scale this dynamic, high-performance storage system to more than half an exabyte (>500 petabytes).

"NCSA and the University of Illinois are creating this environment to enable our science and engineering partners to pursue scientific discovery and innovation. This environment fulfills the research community's tremendous need for big data services," says Michelle Butler, leader of NCSA's Network and Storage Engineering team.

"This near-line storage acquisition is part of NCSA's strategy to broadly support data-intensive research," says NCSA Chief Technology Officer and deputy director Rob Pennington.

NCSA will initially deploy four Spectra Logic 17-frame T-Finity tape libraries, which will hold a total 244 IBM TS1140 tape drives and will house more than 19,000 media slots in each library. In 2013, NCSA will deploy two more libraries with an additional 122 tape drives. This environment will provide 380 petabytes of raw data slots. Each tape drive's performance is 240 megabytes per second.

By the end of year one operations for the Blue Waters supercomputer, NCSA projects that the system will hold 50 petabytes of science and engineering data—more than any other U.S. research institution.

NCSA will use a RAIT (Redunant Array of Independent Tapes) technology written by NCSA and the HPSS consortium—a new feature of HPSS that provides data integrity/protection through parity technologies for the tape subsystem. "At this scale, we can't afford a mirror copy, but a redundant array of tapes with eight data and two parity can survive the loss of two tapes at a cost of only 25 percent more, which means that through software development and engineering we can still afford to provide data integrity," Butler explains.

National Science Foundation

Blue Waters is supported by the National Science Foundation through awards ACI-0725070 and ACI-1238993.