Communication latency hiding - model and implementation in high-latency computer networks
Volker Strumpen · Repository for Publications and Research Data (ETH Zurich) · 1994
The potential of large numbers of workstations for solving very large problems is tremendous.Nevertheless, it is often considered inappropriate to parallelize applications with a fair amount o f communication on computer networks, because communication via networks with high latency and low bandwidth presents a technological bottleneck.In this paper, a model to analyze the gain of communication latency hiding by o verlapping computation and communication is described.This model captures the limitations and illustrates the opportunities of communication latency hiding for improving speedup and eciency of parallel computations that can be structured appropriately.F urthermore, an implementation of a message passing protocol is presented that incorporates latency hiding on top of the TCP/IP transport layer.This protocol ensures ecient, deadlock-free communication in UNIX network environments.Experiments show that the presented latency hiding technique increases the range of applications suited for parallel computing on networks of computers, even across the Internet.Measurements with a multiprocessor system demonstrate the validity o f the latency hiding model for a broader range of parallel architectures.Parallel programming with conventional message passing interfaces is only slightly aected, because an additional protocol layer hides the increased complexity from the programmer.