Title: Or-Parallel Scheduling Strategies Revisited 
Authors: Ines de Castro Dutta and Adriana M. Carrusca

Presenter: Hung Viet Le

Abstract:

Parallel logic programming systems have been studied for more than
a decade. Techniques and scheduling strategies for or-parallel
systems have been established for systems that run on centralised memory
architectures. As new parallel platforms such as clusters of workstations
or clusters of PCs gain popularity, these techniques vastly studied
for centralised memory systems may become obsolete and inefficient. In
this work we study several scheduling strategies commonly used in
orparallel systems designed for centralised memory architetcures. We
simulate these strategies on different parallel environments in order
to estimate the costs associated to these scheduling strategies in each
environment. We use a benchmark set commonly studied by the or-parallel
community. Our study concentrates on simulating top-most, bottommost,
and left-most strategies while modelling costs associated to
centralised and distributed memory architectures. We then implement our
own strategy that selects best work based on the costs to move to a piece of
work.