Research On Resource Dynamic Management Based On SLA In IaaS Cloud Computing Environment

With the development of information technology in our society, after the three big computing models--Parallel computing, Distributed computing and Grid computing were proposed and developed in particular level, a new computing model, called Cloud computing was taken out by scholars. Through the Internet, Cloud computing can integrate computing resources and storage resources into a big resources pool and provide its customers the ability of on-demand and scalable computing functionalities. Customers request certain resources towards data center and sign Service Level Agreement (SLA) with the provider to ensure their published applications can always be accessed quickly without serious delay. However, the quantities of resources used by customers’applications are changed with the change of the time, showing some certain regularity. If providers provide constant number of resources with the static resources scheduling strategy, the Service Level Agreement can be assured, but when the service is idle, the allocated resources will be wasted a lot. So, how to guarantee services’ SLA and schedule the resources in high efficiency is to be solved in the Cloud computing center and is also our problem to be solved in this paper.The appearance of virtualization technology make us to manage the center of Cloud computing being possible with dynamic resources scheduling strategy. It allows many virtual machines running in the same physical in the same time. These virtual machines do not affect each other. The applications are encapsulated into the virtual machines and then published to the corresponding physical machine. But virtual machines can be live migrated from one physical machine to the other physical machine. In this process, the running applications will not be influenced. So, the resources providers can integrate resources through live migration technology, making virtual machines are distributed on less physical machines. Then the physical machines without any virtual machine will be shut down or being low power consumption to save energy cost or serve for more customers. Particularly, we can largely lower the energy cost in data center and reduce more environmental pollution through dynamic resources scheduling strategy with virtualization technologyThrough many researches and analyze in this field, we provide a heuristic dynamic resources scheduling strategy. This strategy can reduce the times of virtual machines’ migration because of needing adjusting the resources allocation. It can not only reduce the number of active physical machines in the largest extent, but also lower the probability of violate SLA. To prove the validity and robust of our strategy in this paper, firstly, we did a experiment in the form of simulating with the tool--CloudSim. Compared with MM and ST strategy in the case of different numbers of physical machine, our strategy has better expectation in ensuring SLA and reducing the energy cost in the aspect of the times of virtual machines’migration and the probability of violation of SLA. At last, we use the open source and virtualization software Proxmox to create a small private cloud computing platform and then do a secondary development based on the platform. We integrate our strategy into this platform and implement a real cloud computing resource management system. After many tests, we can see that this prototype system can not only implement auto-management perfectly, but also prove its availability in the actual situation.