Research On Bandwidth Allocation In Cloud Computing Datacenter

In recent years, the way of people’s obtaining resource is changing owing to development of Cloud Computing. With the support of Cloud Computing, Many IT enterprises choose to deploy their service into the cloud and provide their everywhere service access to customers conveniently. As the carrying entity of Cloud Computing, Datacenter manages many resource s, such as computing resource, storage resource and bandwidth resource. As for the resource of computing and storage, datac enter allocate them according to the payment of their customer, namely “pay-as-you-go”. Different with the computing and storage resource, the bandwidth is allocated in a “best-effort” way, which does not provide the isolation in the level of tenant or virtual machine and further gives rise to the interference among different tenants and virtual machines. On the other hand, the bandwidth demand of tenants in datacenter is variable and dynamic, which brings the problem of bandwidth churn caused by bandwidth reallocation. This also does harm to the performance of network in datacenter.To handle these circumstances above, som e solution schemes try to promote the performance of tenants or virtual machines from two aspects. One chooses to guarantee the basic bandwidth demand of virtual machines and the other allocates the bandwidth to tenants or virtual machines in proportion. Based on these researches, the contributions of our paper focus on the following three aspects:(1)On the basis of investigation on the policy of bandwidth allocation in cloud computing data center, we analysis and summarize these strategies and point out that they lack the limitation on the bandwidth churn because of bandwidth reallocation.(2)To handle the unpredictable bandwidth performance and bandwidth churn problem, the parameter of bandwidth satisfaction and control of bandwidth churn are proposed to help build the mod el of bandwidth allocation in datacenter, which decreases the extent of bandwidth churn from 10% to 5%.(3)To achieve the solution of the model above, we proposed the multi-stage bandwidth allocation algorithm based on the inequality of arithmetic and geometric means, which can decrease the time complexity of resolving problem from O(2) to O(nlogn).