Node Availability for Distributed Systems considering processor and RAM utilization for Load Balancing

Abstract

Node-Availability is a new metric that based on processor utilization, free RAM and number of processes queued at a node, compares different workload levels of the nodes participating in a distributed system. Dynamic scheduling and Load-Balancing in distributed systems can be achieved through the Node-Availability metric as decision criterion, even without previously knowing the execution time of the processes, nor other information about them such as process communication requirements. This paper also presents a case study which shows that the metric is feasible to implement in conjunction with a dynamic Load-Balancing algorithm, obtaining an acceptable performance.