Research On Two-Level Resource Scheduling Algorithm Based On Cloud Stack

The traditional IT infrastructure has been changing rapidly by Cloud Computing due to the unique business model of Cloud Computing, which arouse a wide spread attention in IT industry. With the continuously enrichment of public cloud services and development of both private cloud and hybrid cloud, the size and quantity of cloud data centers are constantly expanding. How to effectively manage the resources of cloud data center is a challenge to all kinds of Cloud Computing platform. CloudStack is a complete open-source Cloud Computing platform, which is not only one of the largest leading open-source projects in the Apache but also has many unique advantages that attract numerous commercial customers. The advantages include the compatibility with most mainstream hypervisor, the support of a large number of storage types, the enhanced ability to interact with the network, abundant UI features and so on. However, there exist many deficiencies in CloudStack’s virtual machine deployment strategy and its resource rebalancing method. For instance, when firstly using matching algorithm during the virtual machine deployment, CloudStack does not take the performance parameters of the host and other virtual machines into consideration. Furthermore, when the virtual machine deployment is done, there is no more procedure in CloudStack to reschedule the resources during the process of operation. To solve the load balancing problem of cloud data center, the study of this thesis is mainly on the following aspects:(1) We study the corresponding techniques about Cloud Computing and the mainstream of the open source cloud platform, especially analyze deeply the cloud platform of CloudStack, and its resource scheduling strategy has been carried out. Based on those analysis, we point out that the drawbacks of the existing resource scheduling strategy are insufficient when scheduling the resource in different production environment.(2) We propose the two level resource scheduling algorithm based on the integrated load rate, which improve the shortcomings of the existing methods. And meanwhile exploit the implementation of this algorithm, which provide a reference for later cloud resource scheduling research.(3) We develop a simulation system of the two level resource scheduling algorithm based on CloudStack, which can be used as an experimental tool for the simulation of the new algorithm and the existing strategy.(4) By means of the simulation system, we do some detailed experiments to the existing resource scheduling strategy and the two level resource scheduling algorithms of CloudStack.The above research and analysis show that the CloudStack-based two level resource scheduling algorithm is a feasible and effective method to load balance and manage the resource of cloud data center.