Network and Memory Analysis in Distributed Parallel Generation of Pat Arrays
João Paulo W. Kitajima, Berthier Ribeiro, Nívio Ziviani · 1996
The performance of parallel and distributed algorithms for generation of large pat arrays is analyzed. These algorithms are evaluated taking in to account a high-bandwidth network of workstations, a TCP/IP-based network and an heterogeneous network with different memory sizes. In the first case, performance of the parallel versions are significantly better. In the second case, the sequential algorithm is clearly the best. In the third case, different memory sizes will hardly improve execution times significantly.