Distributed computing systems may be constructed from various commodity computing resources in an organization in order to exploit otherwise idle processing time. When such a system is used to solve embarrassingly parallel problems, the performance of each computing resource in the system becomes a significant factor in overall processing performance. This paper shows that processing performance in a distributed computing system used to process tasks of equal duration declines as individual processing resources diverge with respect to performance. A new dynamic load balancing algorithm based on the division of labour in ant colonies is proposed and its viability tested in terms of the time taken to complete a distributed computation and the bandwidth utilized within the system.
展开▼