Distributed image processing on a network of workstations

Concurrent computing on networks of distributed computers has gained tremendous attention and popularity in recent years. In this thesis, we study and implement various data parallel image processing applications on a network of workstations. Experiments with different image sizes, mask sizes and number of workstations are carried out. Significant speedup is achieved in this computing environment and compares favorably to traditional multiprocessor systems. We also present a performance prediction model that agrees well with our experimental results and allows the highest speedup to be estimated from the knowledge of the ratio of the computation time to the communication time. A dynamic load balancing strategy is also proposed to tackle the problem of imbalanced workload among workstations. Simulation and experimental studies of our load balancing strategy are performed for different load situation and show that it can effectively balance the workload among workstations, particularly under a heavily loaded system, to reduce the execution time of the parallel image processing applications. The main limiting factor in our computing environment is the bandwidth of the network. Thus, it seems with emerging high-speed networks, parallel computing on networks of distributed computers can be a very attractive alternative to traditional parallel computers