Research On Multi-dimensional Attribute-driven Resource Availability In Cloud Environment

With the development of network bandwidth, the condition of accessing remote service(for example, data processing, storage and information service) through network has become more and more mature, and then cloud computing appears. It integrates grid computing, virtualization and SOA etc. together while gathering storage, platform, application and computing capability which is abstract, virtual and scalable into resource pool, to provide service to users according to their needs. Users have no knowledge of technologies details at all, but to access the cloud through interfaces. As cloud environment is highly commercialized and user-central, it is important for researchers to solve the problem of providing service of high scalability, high availability, high security and high efficiency. However, at the same time, the complexity of cloud is blocking in the way.Resource Management is the core component in cloud environment. The goal of resource management in cloud is to make use of resources effectively and reasonably, which can acquire high quality, efficiency and profit at less cost. There are two major management difficulties here: one is how to make full use of historic and current information, to achieve effective resource scheduling to meet the performance and QoS needs of users; the other is data management ensuring high availability of data access service at the premise of gaining efficiency and consistency.Focusing on these two issues of resource scheduling and data management, this thesis carries out researches as follows:1. Investigate the research background, including the concept, features and applications. State the process and background of cloud computing. Analyze the differences and connections between grid computing and cloud computing. Then simply introduce current related researches and propose the research topics.2. Introduce the content of resource management in cloud environment. Summarize the resource types and analyze the key techniques of resource management which involve resource scheduling, data management and virtual machine, energy consumption management etc. Then discuss the problems facing in front of the techniques.3. Comprehensively consider the rate of CPU usage, level of network load, and resource failure to forecast resource future state for getting better job scheduling results. An evaluation measurement is designed to quantify the prediction result for scheduling. Experiments show that this method has a better performance on resource idle rate and prediction accuracy.4. Availability is the key element in cloud storage. The complexity of cloud storage brings availability measurement to be a difficult problem. This thesis suggests an availability measurement method, which simplifies the environment, and proposes the definition of availability and computing method based on updating failure and node failure, then illustrates with examples.By analyzing the concept and characteristics of the cloud resource management and integrating a variety of resource properties, this thesis researches into resource availability in the cloud environment, and proposes a resource prediction mechanism and an availability method by introducing queuing theory and Markov model. The work contains two parts: the first part predicts the resource availability in the future, guiding resource scheduling to improve its efficiency and resource utilization; the second part analyzes the impact that system parameters produce on resource availability, and provides reference for data management strategy for optimization of the system QoS.