Optimization of probabilistic & decentralized load balancer on SSI cluster

The work done in this research attempts to improve the load balancer of OpenMosix. Interestingly trivial changes made to any load balancing algorithm can produce dramatic results. The parameters that effect the load balancing algorithm are first evaluated with the default configurations and then modifications have been made to the /proc file-system dynamically to optimize it in OM. A stronger system in the SSI is assigned most of the workload for a given process whereas relatively slower systems are not as much active; hence reducing the overall efficiency. Making the faster systems pretend to be slower, results are more optimized among the whole SSI nodes. The results show that tuning the speed and decay interval parameters from the /proc interface improves the performance of the overall SSI for the given task.