Developing a Framework for Analyzing Data Movement within a Memory Management Runtime for Data-Intensive Applications




Abstract: We present a framework for data-intensive runtimes that provides online data movement and caching behavior information to an application. This framework is built on top of the data-intensive memory-map (DI-MMAP) runtime and helps designers observe the efficiency of their I/O scheduling and concurrency. Furthermore, we propose a lightweight mechanism for tracking page cache occupancy to enable applications to make online, dynamic scheduling decisions that maximize data reuse.

Bio: Brian Van Essen has been a Computer Scientist at Lawrence Livermore National Laboratory (LLNL) since 2010. His research interests include operating systems and architectures for data-intensive HPC, deep learning, and embedded systems. Brian earned his Ph.D. in Computer Science and Engineering (CSE) from the University of Washington in Seattle in 2010. He also holds a M.S. in CSE from UW, plus a M.S. and a B.S. in Electrical and Computer Engineering (ECE) from Carnegie Mellon University.