Energy-aware And Reliable Cloud Computing Task Scheduling

With the widespread use of cloud services,cloud service providers need to continuously improve Quality of Service(Qo S)and reduce operating costs.The extreme complexity of cloud data centers causes frequent hardware and software failures,which in turn leads to huge losses.However,fault-tolerant mechanisms to deal with hardware and software failures will inevitably increase the cost of cloud computing systems,including energy consumption.Therefore,quality of service,energy consumption and reliability have become the focus of cloud service providers and users.In response to the above problems,there is an urgent need for efficient task deployment strategies to deploy tasks to appropriate cloud data center servers to improve the reliability,quality of service,and energy efficiency of the cloud data center.The main research contents of this dissertation are as follows:(1)An energy-aware and reliable cloud computing task deployment algorithm is proposed.At present,most researches on cloud computing task deployment only focus on one of the two goals of reliability and energy consumption.This article studies how to provide fault tolerance for task execution failures while minimizing the number of servers used to execute all task.Thereby reducing the problem of energy consumption.This article provides fault resilience by task replication,that is,each task composing the job has two instances:main task and task replica.Among them,each main task runs at full speed on a separate server,and the task replica associated with the main task run on a different server.To reduce the energy consumption,the task replicas run at a lower rate than the main tasks by sharing the same computing resources,and the task replicas can be deployed either on a dedicated backup servers or on the server where the main task is located.On this basis,this dissertation proposes a Reliability-aware and Energy-efficient task Replicas deployment algorithm(RER).The algorithm uses low-speed task copies and allows multiple task copies to share the same server resources to make full use of each server.Let ?^max and ?^min denote the maximum and second minimum values of the idle period time on the server,respectively,?=?^max/?^min.Theoretical analysis proves that the algorithm RER is an approximate algorithm with an approximate ratio of 3?/2.The simulation results show that the algorithm can effectively balance between energy consumption and job completion time.(2)A reliable cloud computing task deployment algorithm for energy consumption optimization and Qo S-aware is proposed.For users,there is usually a certain limit to the completion time of the service.If the completion time is out of time,it will cause losses to the cloud service provider.This dissertation studies how to minimize the number of servers used to execute copies of all tasks while ensuring quality of service and providing fault tolerance,thereby reducing energy consumption.This article deploys task replicas of different sizes for tasks of different sizes and deploys them to a dedicated backup server to execute at different execution speeds to ensure that the task copies can still be completed within the time required by the user after a failure occurs.On this basis,this dissertation proposes a Qo S-aware,Reliability,and Energy-efficient task replicas deployment algorithm(QSRE).Theoretical analysis proves that the algorithm QSRE is an approximate algorithm with an approximate ratio of 3/2.The simulation results show that the algorithm effectively reduces the number of servers used under the premise of ensuring reliability and quality of service,thereby reducing energy consumption.