Robustness analysis of a new distributed scheduling approach - Relprox model using Mosix

This paper presents a new distributed scheduling approach, Relprox model for a trusted small scale distributed computing environment. The main goal is to determine which process to migrate, where to migrate and when. Our algorithm uses the information provided by the Load Information Management(LIM) module, Proximity Cluster Table(PCT), Node Reliability Table(NRT) and Process Priority Table(PPT) to make migration decisions. Process Migration will be carried out between client nodes without it being sent to server first. This algorithm not only reduces the latency time in migration and execution, but also reduces network overhead. Our algorithm proposes two new concepts of reliability and proximity. The reliability check in our algorithm ensures transferring of process to a stable node resulting in minimum number of migrations and hence improving efficiency of the whole system. Proximity ensures transferring of work load to the closest possible node that reduces network overhead and execution time. To implement this mechanism, we use the hybrid architecture and Relprox algorithm with Mosix as a middleware. This model is most suitable for running High Performance Computing(HPC) applications with low to moderate amount of I/O.