home · mobile · calendar · colloquia · 1999-2000 · 

Colloquium - Bennett

Brazos: High Availability Parallel Computing on Multiprocessor Clusters
Rice University

Clusters of industry-standard multiprocessors are emerging as an attractive alternative for large-scale parallel computing. Improvements over the last decade in commodity general-purpose processors and networks, and in particular the development of small-scale symmetric multiprocessors, have made feasible an inexpensive alternative to large bus-based multiprocessor systems.

This talk will describe Brazos, a parallel programming environment for shared memory and message passing applications on clusters of SMP workstations. Brazos implements a multithreaded shared virtual memory system using virtual memory protection mechanisms to maintain coherence across multiple nodes, and relies on hardware cache-coherence to maintain coherence between user threads in the same process. Brazos makes selective use of multicast communication, adaptive runtime performance tuning, and a software adaptation of scope consistency.

Multiprocessor clusters also exhibit several disadvantages relative to large monolithic systems, including complex thread scheduling and increased susceptibility to failure. Brazos implements two user-level mechanisms that address these issues: thread migration and checkpoint/recovery. The implementation and performance of the migration and checkpoint/recovery mechanisms in Brazos will be described.

Bennett is a candidate for a faculty position in the Department.
Refreshments will be served in ECOT 831 immediately following the talk.
Hosted by Gary Nutt.

Department of Computer Science
University of Colorado Boulder
Boulder, CO 80309-0430 USA
May 5, 2012 (14:13)