Research On Energy-efficiency Aware Resource Scheduling Algorithm Based On Customer Satisfaction Level

With the tremendous investment in cloud data center infrastructure around the world,high energy consumption in cloud data centers has gradually become a prominent problem,resulting in high carbon emissions and environmental pollution.The rapid development of virtualization technology provides a solution to the energy management in cloud data centers.The dynamic consolidation of virtual resources based on energy efficiency model greatly improved the energy efficiency of the data center and reduced the cost of cloud service.Customer Satisfaction Level(CSL)and energy efficiency are two main objectives.A trade-off model based on the number of SLA violations and power consumption is designed to improve energy efficiency.CSL could be classified into three distinct stats according to success rate:imperceptible,tolerable,and unusable region.A prediction algorithm is introduced to decide which region it belongs to.As a result,the proposed trade-off model and prediction algorithm greatly improve CSL per energy.Our contributions in this thesis are:(1)A trade-off model is designed that trades off CSL and power consumption.An energy-efficient resources management not only takes into account the energy-efficient model and consolidation algorithm of virtual resource,but also meets the requirements of Customer Satisfaction Level(CSL).However,there exist constraint relationship between CSL and power consumption.It is important to improve energy efficiency under Customer Satisfaction Level constraints in cloud data center.Compared with power model that focuses on minimizing power consumption,a CSL-aware model is proposed to improve Customer Satisfaction Level in cloud computing.To capture the trade-off between CSL and power consumption,we proposed a trade-off model which aims to maximize the CSL per power consumption.(2)Propose a prediction mix model called WMA based on success rate.The Customer Satisfaction Level is defined in terms of a variety of properties such as response time,throughput,cost,reliability,availability and etc.A Service Level Agreements(SLA)is defined as an official commitment that prevails between a service provider and the customer to ensure that their Customer Satisfaction Level(CSL)requirements are met.CSL has been divided into three region using success rate: imperceptible,tolerable,and unusable region.A prediction algorithm based on Weighted Moving Average is introduced to decide which region it belongs to.This mechanism improve CSL per energy while the overhead could be totally ignored.(3)Compare trade-off model and WMA prediction mechanism with the traditionalalgorithms.This thesis validates the advantages of the proposed trade-off model by analyzing the SLA violations rate of response time and throughput.CloudSim is an open source cloud computing simulator developed by Rajkumar Buyya to support for modelling of large scale cloud computing infrastructure.We use CloudSim to simulate resources consolidation to test the effect of above models.The performance of the three models based on DPSO algorithm is investigated by a series of indicators such as energy consumption,the number of physical nodes enabled and the number of virtual machine migration.WMA prediction mechanism reduces SLA violations rate and the number of host overloads.The experimental results show that proposed trade-off model and WMA prediction mechanism are better than the power model and the CSL-aware model,effectively improve the CSL per power consumption in cloud data center.