Parallelizing a New Class of Large Applications over High-speed Networks

Abstract

Several large applicationshave been paralleli,zed on Nectar, a network-based multicomputer recently developed by Carnegie Mellon.These applications were previously either too large or too complex to be easily implemented on distributed memory parallel systems.Parallelizing these applications was made possible by the cooperative use of many existing general-purpose computers over high-speed networks, and by an implementation methodology based on a clean separation between applicatiionspecific and system-specific code.We illustrate these points using our experience with parallelizing three real-world applications.The success in these applications clearly points out a new direction in parallel processing.The Nectar system [1] developed by Carnegie Mellon is intended to provide general support for parallelizing large applications.The system is a multicomputer built around a high-speed network.The use of existing general-purpose computers as its nodes and the highbandwidth and low-latency network makes the system inherently suited for large applications.The system has allowed us to parallelize applications that were previously either tm complex or too communicationintensive to be suited for parallel processing.This paper describes the Nectar implementation of three applications: (1) COSMOS [4], a switch-level circuit simulator developed by Randy Bryant and his associates at Carnegie Mellon; (2) NOODLES [7], a solid modeling package developed by Professor