Authors : R. Reshmi and D. Shanthi
Abstract: A scheduler has to assign the tasks to heterogeneously distributed computing sites to process with the objective to achieve the customers’ Quality of Service (QoS) requirements as well as to optimize the performance. In this study, a QoS oriented mechanism, evaluates nodes in a high performance cluster based on various static and real-time parameters. It selects the most appropriate node to perform the task at hand with regard to their quality of service. The nodes are gauged on parameters, spanning a broad spectrum from power usage to fan speed. This system offers an inherent flexibility as it can be tailored to bias the selection towards a particular parameter or can be maintained on a balanced plane. In particular, the focus was on designing a power aware system, where power usage, CPU temperature etc. play a critical role in determining the QoS of a node. The result of our experiments that upholds the optimality of QoS based selection algorithm is also presented.
R. Reshmi and D. Shanthi, 2016. QoS Oriented Node Selection in High Performance Clusters. Asian Journal of Information Technology, 15: 1307-1313.