Two memory interconnection networks called Omega and Crossbar were compared using simulation. Two different memory address reference patterns called Random and, Matrix were used. As the name implies, Random reference pattern addresses the memory with a uniform probability of reference. The second pattern simulated a matrix multiplication problem in which each processor (of a multiprocessor system) is doing a part of the multiplication. Tb keep the analysis simple, many factors that were known to affect the performance of the interconnection networks were kept fixed at one level as follows:
1. Number of processors was fixed at 16.
2. Queued requests were not buffered but blocked.
3. Circuit switching was used instead of packet switching.
4. Random arbitration was used instead of round robin.
5. Infinite interleaving of memory was used so that there was no memory bank contention.