A Review of “ Customized Dynamic Load Balancing for a Network of Workstations ” Taken from work done

With the rapid advances in computational power, speed, cost, memory, and high-speed network technologies, a “Network of Workstations” provides an attractive scalable alternative compared to custom parallel machines. Scalable distributed shared-memory architectures rely on the ability for each processor to maintain a load balance. Load balancing involves assigning each processor a proportional amount of work bases on its performance, thereby minimizing total execution time of the program. There are two types of load balancing, static and dynamic. Static load balancing allows the programmer to delegate responsibility to each processor before run time. The simplest approach is the static block-scheduling scheme, which assigns equal blocks of iterations to each of the processors. Another scheme of static load balancing is the static interleaved scheme, which assigns iteration in a cyclic fashion. This makes programming much easier and can solve many problems inherit in load balancing caused by processor heterogeneity and non-uniform loops. Static scheduling avoids run time scheduling overheads.