On the allocations of documents in the distributed memory information retrieval system

A proper document allocation in a distributed memory information retrieval system is needed to improve the efficiency and the effectiveness of such a system. The allocation has to minimize the interprocessor communication traffic, at the same time maximizing the amount of parallelism. These requirements are NP-complete problems, and are conflicting. Genetic algorithms (GAs) are used to find an optimal allocation, and proved to be satisfactory. An improvement to a particular GA is successfully applied to refine the solution. During the experimentation, an interaction between the GA population size and its mutation probability is revealed.<
>