A simulated annealing approach to distributed file and task placements

In a distributed system, to find the placement of files and tasks at the sites with minimal total communication overhead is consequential. To this end, a genetic algorithm (GA) has been developed. However, the operations involved are fairly complicated and time consuming. Besides, considerations for the problem's ''objective'' are not practical enough. For improvement, we propose the adoption of the simulated annealing (SA) approach and the use of multiple objective functions to achieve desirable solutions for the problem. Extensive simulation runs are conducted to collect the results produced from both the GA and SA approaches for various data sets. The SA approach is shown through experimental results to depict superior performance in obtaining file and task placements, with much less complexity.