Abstract
Linear and mixed programming seem like ideal candidates for massively parallel processing. Such problems have enormous appetites for computing resources that grow quickly as the amount of available data expands, the results of these computations have tremendous strategic and financial value, and the algorithms commonly used to solve such problems provide seemingly obvious sources of significant parallelism. Despite all this, large-scale parallel computing plays only a modest role in practical optimization. We'll look at some of the algorithmic challenges that stand in the way of more widespread adoption of parallel computing.