2012 Early Career Academic Achievement Alumni Award
An associate professor of computer science at North Carolina State University, Xiaosong Ma (PhD CS ’03) is conducting research in storage systems, parallel input/output (I/O), high-performance parallel applications, cloud computing, and self-configurable performance optimizations. Ma is also a faculty member at Oak Ridge National Laboratory, giving her access to supercomputing resources and allowing her to collaborate on projects that explore petascale supercomputer management, scalable parallel I/O, and high-end storage systems.
I/O is a weak link in the scientific computing cycle for many applications because of the widening performance gap between the I/O subsystem and other system components. Ma’s group works to provide novel technology for improving applications’ perceived I/O performance as well as for reducing parallel jobs’ data movement cost in time, resource usage, and energy consumption. More recently her research is exploring the use of new storage devices such as SSDs in HPC settings for efficient out-of-core computation and in-situ data analytics using active-SSDs. Another recent focus is automatic parallel I/O benchmark extraction based on large-scale applications.
Another aspect of her research is focused on making cloud computing a cost-effective choice for HPC users. Ma's group has recently investigated the cost comparison between using cloud instances and owning in-house clusters for executing tightly-coupled parallel applications. They are also examining new approaches such as establishing semi-elastic, cloud-based clusters and automated per-application parallel I/O configuration on cloud platforms.
Ma has received a Department of Energy Early Career award, a National Science Foundation CAREER award, an IBM Faculty Award, and a Faculty Fellowship from NetApp.