Dynamic load distribution methods are necessary for parallel processing of unstructured computation based on a message-passing multiprocessor system. To get faster response time, the loads should be distributed rapidly and uniformly to all processing elements as soon as possible.
In this thesis, we suggest a load distribution algorithm, Relative Load Comparison (RLC), and compare it's performance with that of algorithm Contracting Within Neighborhood (CWN) proposed by L.V. Kale at University of Illinois. A comparison is made by the extensive simulation.
The simulation results show that our Relative Load Comparison (RLC) load distribution algorithm is more effective than the Contracting Within Neighborhood (CWN) algorithm.