An object oriented approach to algorithm management on networked workstations

This paper presents the results from the use of a parallel algorithm manager which automatically configures and manages networked workstations for the solution of optimization problems. The algorithm manager has an object oriented interface, resulting in an easy to use platform to perform parallel computational experiments. Network programming protocols are encapsulated within the algorithm manager to minimize errors associated with misuse. Results from a parallel branch and bound fixed charge problem are presented.