Energy Efficient Utilization of Computational Resources by Predicting CPU Idle Time
Ramesh T1, R M Suresh2
1Ramesh T, Sathyabama University Chennai (TamilNadu), India.
2Dr R M Suresh, Sri Lakshmi Ammal Engineering College, Chennai (TamilNadu), India.
Manuscript received on 10 April 2019 | Revised Manuscript received on 17 April 2019 | Manuscript Published on 24 May 2019 | PP: 66-70 | Volume-8 Issue-6S3 April 2019 | Retrieval Number: F10130486S419/19©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open-access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Consumption of energy in cloud data centers are growing at an alarming rate owing to regularizing of using cloud servers for data and resource sharing applications. The consumption of power usage in CPUs is not directly proportional to its utilization as the power feed to the cloud server is not regulated as per its CPU utilization levels. Hence, increasing the operational and maintenance cost. Though this problem can be curved if we could predict the idle CPS utilization time of variable lengths and thereby regulate its operational mode into various low-power or sleep mode in order to automatically control the power utilization. But such a system would require to synchronously train on the live CPU data as the problem of waking up the CPU at unoptimized level would result into latency issues and thereby reversing the outcome causing more power wastage then actually destined to save so. Therefore, development of such a online learning framework is a significant challenge for achieving energy savings at data centers. In this study, we present aonline learning based algorithm for CPU idle state prediction by training it on real time over the live streaming data of Cloud servers and its state of CPUs. Our results show that about 50-60% of power savings can be achieved over variety of other existing techniques for prediction CPU idle times.
Keywords: Cloud Computing, CPU Ideal Time, Workflows, Management, Online Learning.
Scope of the Article: Mechanical Maintenance