NOWs (Networks of workstations) have been extensively used
to execute parallel applications. Although NOWs seem to be an easy
and inexpensive way of obtaining great performance, it may not
always be so. When using such an environment for executing a
parallel application, performance may not be as good as expected due
to delays in communication. Also, the heterogeneity in communication
makes it hard to take advantage, or reuse, communication strategies
that were useful in regular-topology platforms, e.g., parallel
machines or LAN-based clusters of workstations. For instance,
broadcasting in a NOW may be more challenging due to the variety of
communication links and, consequently, of point-to-point latencies.
In this paper, we present a strategy to improve broadcast algorithms
that are based on binomial trees, which are used in regular-topology
platforms, so that they can execute efficiently in Local NOWs.
|