Genetic Algorithm Based Process Migration to Optimize Downtime

Abstract

A prominent research area that has gained attention recently in networks, newlinedistributed systems and cloud computing is load balancing. Distributed operating system newlinehas become an intriguing topic for research in the current scenario and newlineachieving/implementing load balancing is challenging in distributed operating system. newlineTraditionally, process migration mechanisms have been used to balance load on newlineprocessors in distributed systems and approaches for supporting them transparent to the newlineapplication has required extensive kernel support. newlineThe term load balancing refers to any movement of workload from one computer newlineto another for the purposes of improving performance. Load balancing in distributed newlinesystem therefore refers to transmitting processes between nodes i.e., from nodes with newlinehigher loads to nodes with lighter/moderate loads. Load balancing hence aims at newlinedecreasing the average time required to complete processes and improves processor newlineutilization. The strategy to balance load includes two important aspects namely migration newlinepolicy and location policy. The decision as to when to migrate and which process needs newlineto be migrated is taken care of by the migration policy. Whereas location policy intends newlineto pick the destination node/victim node on to which the process would be migrated. The newlineload can be balanced either implicitly by the system or the user can explicitly implement newlineit. The implicit policies may or may not use the prior information of the processes such as newlinethe utilization of resources including CPU i.e., CPU load, memory load and other newlineresources and network conditions. Any approach to balance load needs to consider newlinevarious factors such as load estimation, categorizing load to decide the states of the newlinenodes, system performance, communication among nodes, type of load to be transferred newlineand destination node selection. newlineWhen the system is heavily loaded, determining a victim/destination newlinenode(i.e.,location policy) becomes a challenge as multiple nodes would possess newlineadditional load that has to be offloaded.