Adaptive Threshold Detection Based on Current Demand for Efficient Utilization of Cloud Resources
Junaid Hussain Abro, Chunlin Li, Qaiser Hammad-Ur-Rehman · 2019 IEEE 4th International Conference on Computer and Communication Systems (ICCCS) · 2019
Efficient utilization of Cloud resources is of utmost importance for Cloud Data Centers. The overall objective is to reduce energy consumption without compromising over quality of service. In order to achieve this goal different Virtual Machine (VM) Consolidation techniques have been proposed. Threshold detection for over utilized and underutilized hosts is an important phase of these techniques. In this paper we have proposed an adaptive model based on current load to optimize thresholds for detection of over and underutilized hosts. The proposed model uses Mean utilization of hosts and Standard deviation of VM utilization to calculate an acceptable range of utilization thresholds. An algorithm based on the proposed model is devised and tested through simulation experiments using Cloudsim. Four threshold detection policies namely, Static Threshold (ST), Inter-Quartile Range (IQR), Local Regression (LR), and Median Absolute Deviation (MAD) with Minimum Migration Time (MMT) as VM selection policy are compared with our proposed algorithm. Experiment results validate that the proposed algorithm reduces energy consumption, SLA violations, number of VM migrations and standard deviation of VM utilization in order to optimize resource utilization considerably.