Research On Energy-efficient Task Scheduling Strategy And Performance Optimization In Heterogneous Cloud Environment

In recent years,with the rapid growth of the cloud computing market,the deepening of cloud level on the enterprise,and the continuous expansion of the cloud user scale,the excessive energy consumption in the data centers has become a social consensus.The high energy consumption not only increases the cost of the operation and maintenance for the data centers,but also causes a huge waste of social energy,which seriously restricts the development of the cloud computing.With the vigorous development of 5G technology and the gradual maturity of the industrial Internet,the social demand for the data centers will further increase.Therefore,it is imperative to study the reasonable energy efficiency management strategies for the cloud resource and develop green cloud data centers.The existing cloud resource management strategies tend to ignore the users’ experience by shutting down some physical machines(PM)to reduce the energy consumption of system,or ignore the system energy consumption by providing excessive services to ensure the quality of service(QoS)for the cloud users,and rarely take both into account.In order to overcome the above limitations,aiming to improve the energy saving level of the cloud system and guarantee the QoS of cloud users,we consider the needs of different cloud users to study reasonable dynamic resource energy efficiency management strategies.By comprehensively applying queueing theory,basic economics theory and optimization theory,we carry out research on the energy-efficient task scheduling strategy and performance optimization in the heterogeneous cloud environment.Firstly,for the same type of task requests,based on a cloud architecture containing heterogeneous PMs,we propose a task scheduling strategy with dual-rate adjustment mechanism and synchronous multiple sleep mode.The cloud architecture under the proposed strategy contains two types of PMs: hot PM(always running)and warm PM(synchronous multiple sleep),in which the task scheduling and moderate supply of cloud resource services are performed according to system load changes.In order to further extend the users’ experience and shorten the average waiting time of users in the system,we improve the sleep mode to the semi-sleep mode for the warm PM,and propose a task scheduling strategy with dual-rate adjustment mechanism and synchronous multiple semi-sleep mode.A warm PM in the semi-sleep mode will run at a lower rate rather than not working at all.Secondly,given the difference in the needs of users for the actual cloud environment,we define two types of tasks: real-time tasks and non-real-time tasks.For saving energy while taking into account the QoS of the cloud users,based on a cloud architecture containing heterogeneous PMs,we propose a task scheduling strategy with priority and partially synchronized multiple sleep mode.The cloud architecture under the proposed strategy includes hot PM(always running)and warm PM(partially synchronized multiple sleep).The tasks are scheduled to accept the moderate cloud resource services in line with the needs of the task requests.In order to further improve the QoS of cloud users,we extend the partially synchronized multiple sleep to the partially synchronized multiple semi-sleep,and propose a task scheduling strategy with priority and partially synchronized multiple semi-sleep mode.For the above task scheduling strategies,we respectively establish the two-stage queueing with adaptive service rate and synchronous multiple vacation,the two-stage queueing with adaptive service rate and synchronous multiple working vacation,the two-stage queueing model with priority and partially synchronized multiple vacation and the two-stage queueing model with priority and partially synchronized multiple working vacation.Based on the matrix geometry solution,the steady state probability analysis of the queueing model is carried out,and the numerical results of the system performance measures are given.Through numerical experiments,the influence of the proposed strategy on the system performance measures is analyzed and evaluated.The experimental results show that the introduction of dual-rate adjustment mechanism is effective to reduce the energy consumption of the system.Compared with the sleep mode,the application of the semi-sleep mode can shorten the average waiting time of users in the system and improve the QoS of users.Finally,in order to balance the tradeoff between the average waiting time of task request,the average energy saving rate of system and the blocking rate of system,the system cost function is established based on various system performance measures under different task scheduling strategies.The performance parameters are optimized.On the basis of the traditional Salp Swarm Algorithm(SSA),we propose an Improved Salp Swarm Algorithm(ISSA)by introducing the chaotic mapping and variation perturbation.With the goal of minimizing the system cost function,ISSA is applied to give the optimization scheme of the performance parameters under different task scheduling strategies.The results of comparative experiments show that ISSA is superior to traditional SSA in convergence speed and global search ability.