Dynamic Network Analysis
At PNNL’s CASS-MT, we are researching, developing, and evaluating dynamic network algorithms on the Cray XMT with initial foci on dynamic social networks and dynamic Bayesian networks.
Dynamic networks (DNs) represent dynamic, often times dependent, systems and processes in which nodes and edges may appear and disappear, loads and bandwidths may fluctuate, and services and contents may move across nodes. There typically exist large-scale systems and processes with multiple types of nodes (multi-node) and links (multi-plex). Algorithms are computationally complex and employ graph data structures; they are typically irregular applications with random memory access patterns.
DNs link to dynamic and/or real-time data sources and require high-performance input/output to handle high volumes of dynamic data. We are working to port and implement specific classes of DN algorithms (dynamic social and Bayesian networks) on Cray XMT to test and evaluate XMT unique capabilities.