MIS Speaker's Series: David Lowenthal
1 p.m. to 2 p.m. Feb. 14, 2020
WhereMcClelland Hall 132
David Lowenthal, Professor of Computer Science at the University of Arizona
Title: Mitigating and Eliminating Inter-Job Interference on High-Performance Computing Systems
On most high performance computing platforms, concurrently executing
jobs share network resources. This sharing can lead to inter-job
network interference, which can have a significant negative impact on
the performance of communication-intensive applications. No
satisfactory solution yet exists for mitigating such performance
degradation on systems that allow jobs to share the network for the
sake of higher utilization. In this talk, we first analyze network
congestion caused by multi-job workloads on fat-tree systems and show
that just a few network hotspots are responsible for performance
degradation. Next, we introduce a new routing strategy to mitigate
this degradation, called adaptive flow-aware routing, which in our
experiments results in up to a 46% improvement in job run time when
compared to the default routing algorithm. Finally, we discuss the
difficulties in completely eliminating inter-job network interference.
David Lowenthal is Professor of Computer Science at the University of Arizona, where he has been a faculty member since January 2009. His research is in all aspects of parallel and distributed computing, with a particular focus on performance analysis, performance modeling, and power management. He was Program co-Chair for the inaugural IEEE Green Computing Conference in 2010, General Chair of the ACM International Conference on Supercomputing in 2011, and has served on numerous program committees, including SC, PPOPP, ICS, and IPDPS.
Prior to Arizona, he was on the faculty in Computer Science at the University of Georgia. He holds a B.S. degree in Computer Science and Math from the University of California, Davis and M.S. and Ph.D degrees in Computer Science from the University of Arizona.