A Clustered Virtual Machine Allocation Strategy Based On Periodic Asynchronous Sleep-mode And Performance Research

With the increase in the number of cloud data centers and the expansion for the scale of cloud data centers,the problem of the control for the energy consumption in cloud systems has become increasingly prominent.In order to realize the green cloud computing,in this paper,towards cloud data centers,applying periodic asynchronous sleep-mode to partial virtual machines,we study the clustered cloud virtual machine allocation strategy and the problem of the performance optimization.Firstly,for cloud systems with high energy-saving requirements,a periodic asynchronous sleep-mode,namely Periodic Asynchronous Sleep-Mode(PASM),is introduced for some virtual machines,and a clustered cloud virtual machine allocation strategy based on PASM is proposed.Considering that some virtual machines continue running at a high speed,and the remaining virtual machines switch between sleep state and low-speed running states,a queueing model with dual-rate and partially asynchronous multiple vacations is established.Secondly,in order to further improve the energy-saving level of the system,a periodic asynchronous sleep-mode with a deferred wake-up threshold,namely Periodic Asynchronous Sleep-Mode with Deferred Wake-Up Threshold(DWUT-PASM)is introduced for some virtual machines,and a clustered cloud virtual machine allocation strategy based on DWUT-PASM is proposed.Considering the deferred wake-up mechanism of the sleeping virtual machine,a queueing model with double-rate and partially N-strategy asynchronous multiple vacations is established.Thirdly,for cloud systems that take into account both the response performance and the energy-saving level,a periodic asynchronous sleep-mode with double control of early wake-up threshold and sleep timer is introduced,namely,Periodic Asynchronous SleepMode with Double Control of Early Wake-Up Threshold and Sleep Timer(DCEWUTSTPASM)is introduced for some virtual machines,and a clustered cloud virtual machine allocation strategy based on DCEWUTST-PASM is proposed.Considering the early wakeup mechanism of the sleeping virtual machine,a queueing model with double-rate and partially(N,T)-strategy asynchronous multiple vacations is established.Finally,using the matrix-geometric solution and other methods,the queuing model with partially asynchronous multiple vacations,partially N-strategy asynchronous multiple vacations and partially(N,T)-strategy asynchronous multiple vacations is solved.The clustered cloud virtual machine allocation strategies based on PASM,DWUT-PASM and DCEWUTST-PASM is evaluated from the aspects of the average latency of cloud tasks and the energy saving rate of the system.The theoretical analysis experiments and simulation statistics experiments were carried out to verify the effectiveness of the clustered cloud virtual machine allocation strategies.The system cost function is constructed,the intelligent optimization algorithm is introduced,and the optimization schemes of the strategy parameters such as the dormancy parameter,the deferred wake-up threshold and the early wake-up threshold are given.