Research On IaaS Resource Scheduling Strategy Based On CloudStack

For the Internet, cloud computing appears to be a revolution. It is widely usedbecause of its characteristics of large-scale, virtualization, reliability, cheapness,and so on. Cloud computing can provide three services of IaaS, PaaS and SaaS. Ofwhich, IaaS is the basis of the other two services. We study IaaS to provide betterservice for the others. IaaS service can be provided by building open-sourceplatforms, among them CloudStack is better in many aspects such as usage, easy-deploying, easy-studying, scheduling policy and so on, so we choose this platformas the research object.Based on the analysis of the CloudStack internal resource schedulingmechanism, we find that the two layers of the resources scheduling plays animportant role in the task optimal span and load balancing. For the first layer, thevirtual machine to the physical machine deployment strategy determines theresource utilization and load imbalance; while for the second layer, the task to thevirtual machine allocation strategy determines the optimal span of the taskexecution time. In this paper, we use the Particle Swarm Optimization (PSO) tooptimize the resource scheduling of the two layers. Because the particle swarmalgorithm has the advantage of high precision and fast convergence speed, it canfulfill the requirements of the scheduling and shorten operation time, so it is theideal scheduling algorithm in cloud computing environment. To deal with theproblem of PSO’s premature, Simulated Annealing (SA) is used to optimize PSO.To deal with the two layers in resource scheduling, we present a virtual machinedeployment algorithm based on improved PSO and dual-fitness task schedulingalgorithm based on PSO, and then the CloudSim simulation tool is used to carry outthe simulation, with results showing that the proposed algorithm can effectivelyimprove the optimal span and optimization of load balancing.In order to make the studied algorithm have practical applications, through thestudy of CloudStack’s source code, we find where the code structure of virtualmachine deployment and task scheduling locates in this open source cloud platformframework. Aiming at this problem, we present a method to improve the internalresource scheduling strategy in CloudStack. For the internal task scheduling inCloudStack, a new task scheduling tool is used to upgrade the original way of taskscheduling, and to increase new functions under the premise of not changingoriginal task scheduling function in CloudStack.As a conclusion, we present an algorithm to improve the resource schedulingstrategy of IaaS, use Simulated Annealing Particle Swarm Optimization (SAPSO) in resource scheduling, and then we analyze the CloudStack open source cloudplatform. The result provides a fine technical support to researchers and enterpriseswho use CloudStack in their cloud computing system.