A Parallel Branch and Bound Algorithm for Solving Large Scale Integer Programming Problems

Abstract

Branch and Bound technique is commonly used for intelligent search in fin ding a set of integer solutions within a space of interest. The corresponding binary tree structure provides a natural p arallelism allowing concurrent evaluation of subproblems using parallel computing technology. While the master-worker paradigm is successfully used in many parallel applications as a common framework to implement parallel applications, it has drawbacks when a large number of computing resources are connected via WAN. A supervisor-master-sub-master-worker algorithm has been proposed. From the solved benchmark example this algorithm proved to provide a considerable save of time. Results show that a consistently bette r efficiency can be achieved in solving integer equations, providing reduction of time. The hierarchical supervisor-master-sub-master-worker algorithm sustains good performance revealed from the knapsack problem solved as a benchmark example.