Scheduling of dynamically evolving parallel programs using the genetic approach

Scheduling of dynamically evolving parallel programs in distributed multiprocessor systems, with different interconnection topologies, is the focus of this study. Each parallel program dynamically evolves during execution time and resulted in a tree-like execution structure. Due to this behavior of the parallel programs, dynamic task scheduling is the technique that is applied here. The parallel programs are scheduled to be run on several interconnection topologies. They are uniprocessor Ethernet, multiprocessors Ethernet, ring, mesh, and hypercube. These topologies are chosen due to their popularity in today's multiprocessor systems. Other researchers have proposed many dynamic schedulers. Among these approaches, many employed search techniques such as genetic algorithm, simulated annealing, hillclimbing, and branch-and-bound. Genetic algorithms have been shown to be a promising technique due to its capability in exploring the solution space. Therefore, the four proposed dynamic schedulers here are based on genetic algorithm. These schedulers are considered as centralized approaches due to the exclusive reservation of one processor for their execution.Master of Applied Scienc