NSF funds STAR lab research on data storage systemsMore than 25 percent of all large-application computer crashes occur because their storage systems are vulnerable, so the National Science Foundation is turning to Tennessee Tech University's Ben He and his students for their expertise in storage systems.
He's research focuses on high availability storage systems for large-scale applications that include systems used for long-term weather prediction, biological processes and nanotechnology. These systems generate and analyze terabytes of data (equal to one trillion bytes or 1,000 gigabytes). He says though reinstalling software and restarting a desktop computer is often the answer to small-scale storage problems that just won't work with large-scale applications.
"At this level, user cannot tolerate rebuilding the system," explained He, an electrical and computer engineering associate professor. "It costs too much time and money. Reinstalling a system managing terabytes of data could take months."
He's group is not the first to try to improve these storage systems, but they are originating how to measure and characterize the availability of storage using a systematic approach.
"We want to develop a way to predict how well a system will perform even under faulty conditions," explained He. "Some applications don't always require peak performance. A system running at 60 or 70 percent may be acceptable for many applications that put availability at the highest priority, as long as you have a way to predict and characterize the results.
"For instance, a system that aids in long-term weather predictions, 10 days out or more, has to reference a lot of stored data," said He. "We want to measure the storage performance and availability under several faulty conditions."
He specializes in research on more efficient, reliable data storage. At TTU's Storage Technology and Architecture Research Laboratory, or STAR lab, he introduced a three-level hierarchy of storage that features a small storage level for retrieval of frequently accessed data, an intermediate larger level for cache and an even larger storage area for backup data.
"Data storage plays an essential role in today's high-end computing systems," He said.
He received this most recent three-year, $235,000 grant from NSF for "A Benchmarking Framework for High-Availability Distributed Storage Systems." This is the fourth NSF grant totaling more than $830,000 He has been awarded in his five years at TTU. Five students from four different countries assist He in the STAR lab. He and his team will collaborate with Stephen Scott of the Oak Ridge National Laboratory on this newly funded NSF project.
He received the university's Sigma Xi Research Award in 2005 and the Ralph E. Powe Junior Faculty Enhancement Award from Oak Ridge Associated Universities in 2004.