Dr. John Chandy, assistant professor of Electrical & Computer Engineering, was awarded a three-year, $538,000 grant from the National Science Foundation (NSF) to conduct research in the area of computer storage subsystems. Dr. Chandy will seek to solve the problem of slow access times for network disk storage performance, which limit computer system performance. His work will benefit high performance computing used, for example, in high energy physics, weather modeling, oil exploration, and genome sequencing.
Dr. Chandy explained that while microprocessor performance and network bandwidths have increased at exponential rates in recent decades, raw disk performance in terms of access times has not experienced comparable gains.
“In recent years, developments in object-based storage systems and parallel file systems have demonstrated the ability to scale aggregate throughput for large data transfers in network storage systems,” said Dr. Chandy. Object-based storage systems – which separate data transfer from the metadata information about the data – save time on the input/output (I/O) process and reduce bottlenecks, because the file server need not search through the hierarchy of a file system to locate the data it seeks.
The problem is that some application workloads are incompatible with certain parallel I/O classes and, thus, do not benefit from these recent advances. Among these applications are those dominated by small data transfers and heavy file system transactional usage as well as I/O patterns that are characterized by reduction operations, explained Dr. Chandy.
Dr. Chandy’s research will focus on inserting active elements in the network, the interface between the storage and processor, to achieve scalable performance. “These active storage networks can draw on knowledge of overall data layout as well as the ability to process all data that is retrieved from the actual storage node. At gigabit/s bandwidths, it is unlikely that a processor in the network can handle many “active” operations; therefore, we will incorporate reconfigurable components in the network to perform these operations in hardware.”
He anticipates his investigation will lead to improvements in network storage system performance and thus contribute to improving our nation’s high-end computing university research infrastructure. In particular, he sees benefits for data-intensive applications of national interest such as the genomics and proteomics, weather simulation, and digital libraries.
Dr. Chandy earned his Ph.D. in 1996 from the University of Illinois at Urbana-Champaign. Before joining UConn in 2002, he was Vice President of Engineering at Sigma Storage, a data management and storage company, and Chief Technology Officer at iChange Corp., an online soft-skills training company. Dr. Chandy’s research interests include storage architectures, optimization using parallel algorithms, distributed file systems and reconfigurable architectures.