Maintaining load balancing in a computingcluster is an evident problem in distributed systems andresearch in this field is not new. The challenges in designing theload balancing algorithms are immense. This paper lists some ofthose challenges in the design of CPU load balancing algorithmand provides solutions to some of them. The algorithmconsiders one node in the cluster as the Master Server andanother as the Load Balancer. The master server maintains theCPU and IP information of each machine. The nodes in thecluster send their CPU status and IP information to the masterserver after every 30 seconds. The implementation solves“readers-writers” problem exclusively using sockets. If anumber of requests are sent before the next central databaseupdate, the load balancer selects other less busy nodes in thecluster. This ensures that all nodes are allocated with the newtasks coming from remote systems, thereby maintaining a loadbalance among the CPUs. This implementation is highly faulttolerant and reliable, guaranteeing a high probability of taskcompletion. Results show that this scheme handles taskallocation in much optimized way and with fewer overheads.The implementation can handle CPUs ranging in numbers from1 to 255.Index Terms—CP
Shoaib Jameel, M., Murugananth, M. & Chingtham, T. Singh. (2009). Deploying CPU load balancing in the Linux cluster using non-repetitive CPU selection. International Journal of Computer and Electrical Engineering, 1 (2), 228-234.