Title: Tree-based Overlay Networks Speaker: Dr. Dorian Arnold, University of New Mexico Date/Time: Wednesday, February 25, 2009, 1:00 – 2:00 pm Location: CSRI Building, Room 90 (Sandia NM) Brief Abstract: As high performance computing (HPC) systems continue to increase in size, scalable, reliable computational models become critical. Tree-based overlay networks (TBONs) leverage the logarithmic scaling properties of the tree organization to provide scalable data multicast, data gather, and data aggregation services. In this talk, I describe our use of the tree-based overlay network (TBON) model to address tool and application scalability. In particular, I present MRNet, our TBON prototype, and several example MRNet applications including debugging and profiling tools developed and used at the Lawrence Livermore and the Los Alamos National Laboratories. I will also highlight some novel algorithms we developed for failure recovery in TBON environments and describe current research directions in scalable, autonomic computing. Bio: Dorian Arnold is an assistant professor in Department of Computer Science at University of New Mexico. His research focuses on scalable performance and fault-tolerance in large scale systems with tens of thousands, hundreds of thousands or even millions of cores. Dorian earned his Ph.D. at the University of Wisconsin in 2008, where he developed MRNet with Phil Roth and Barton Miller. Dorian also worked in the Innovative Computing Laboratory at the University of Tennessee, directed by Dr. Jack Dongarra, as technical lead of the NetSolve project from 1999–2001 -- NetSolve won an R&D Top 100 award in 2000. As a student scholar at the Lawrence Livermore National Laboratory in 2006, Dorian (in collaboration with LLNL scientists) developed the Stack Trace Analysis Tool for effectively debugging large scale applications. Dorian received his M.S. and B.S. degrees in Computer Science from the University of Tennessee and Regis University, respectively.CSRI POC: Ron Brightwell, (505) 844-2099 |