Research On Cloud Computing Resource Allocation Algorithm Based On Three-way Decisions

In recent years,cloud computing,as a cheap and highly scalable computing model,has become more and more favored by enterprises and individuals.With the continuous expansion of the scale of cloud data centers,its resource utilization is unbalanced and energy consumption is wasted seriously.In order to achieve dynamic scheduling of computing resources in cloud data centers,virtual machine migration technology has become critical.Traditional virtual machine migration will be judged according to the load of the physical machine,and the conclusion of the judgment is "migration" or "no migration".This two-way decision ignores the dynamic change of the host load,which is prone to excessive migration.The three-way decision is an optimization of the traditional two-way decision.Under the situation of incomplete information,we can provide a third choice: delay decision,which can reduce the excessive migration of virtual machines.This paper mainly applies three-way decision theories to cloud computing resource allocation,and improves the resource utilization rate of cloud data centers under the premise of ensuring user service requirements.Firstly,in order to determine the optimization goal,this paper analyzes the reasons for the energy consumption of the cloud data center,and constructs four models of the cloud data center: host energy consumption model,virtual machine migration energy consumption model,virtual machine migration cost model,and SLA violation model.In order to provide users with reliable services,this paper studies the fluctuation of the host load,uses the ARIMA model to predict the future load of the host,and puts forward the concept of "migration degree" based on the predicted load value,which is used to express the urgency of physical hosts that need to migrate virtual machines.Secondly,according to the degree of migration,this paper proposes CloudComputing Resource Allocation Algorithm Based on Three-way Decisions,3WD-CRA.Based on the energy consumption model of cloud data center,the loss function matrix is constructed,and three-way thresholds are determined.According to the three-way threshold,the physical hosts are divided into three types: source hosts,target hosts,and optimize hosts.The source hosts include overloaded hosts and underloaded hosts,and the optimize hosts include high load host and low load host,the target host is the host that receives the migrated virtual machine.This paper adopts the "TAO" model of three-way decision to take measures for the three types of hosts.In view of the characteristics that the source host needs to migrate the virtual machine immediately,each time the virtual machine with the highest current CPU utilization is selected to migrate to the target host.In order to balance the load of the optimize host,the virtual machines in the high load host are sorted according to the CPU utilization rate from low to high,and the virtual machines are migrated to the low load host in a lowhigh-low manner.In the target host selection,the migrated virtual machines are sorted from high to low in terms of CPU utilization,and each virtual machine is assigned to the host with the lowest CPU utilization after adding virtual machines.Finally,this paper introduces the CloudSim simulation tool and scheduling strategy,and implements the algorithm of this paper in CloudSim.In the comparative experiment,one classic algorithms in CloudSim,three algorithms of the same type,and two swarm intelligence algorithms were selected for comparison from four indicators.The experimental results show that the algorithm in this paper has lower energy consumption and fewer virtual machine migration times.In the SLA violation rate,the algorithm in this paper is slightly higher than the swarm intelligence algorithm,and the algorithm in this paper performs better on the ESV index.