Research And Implementation Of Resource Scheduling Strategy Under Cloud Computing Environment

Now we have entered a new era-the era of big data. There are T levels data will be produced on the network during our daily life, studying and working. Therefore, a new distributed computing model emerges along with the rapid development of Internet technology, which can calculate the data through dynamically scalable virtualized resources. It solves the issues of massive data storage and real-time backup, and so much for issues that closely related of big data.In recent years, most famous Internet companies and domestic scholars and experts who are all actively engaged the research of cloud computing at home and abroad, and the resource scheduling is one of the key technologies of research. Pros and cons of resource scheduling strategies directly affect the performance of the whole system in cloud computing cluster. Therefore, it has a great significance to the study of resource scheduling strategy in cloud computing environments.The main work of the thesis can be summarized as follows:(1) The paper first introduce the analysis of resource scheduling strategy under cloud computing environment briefly, and study the framework MapReduce and the distributed file system HDFS of Hadoop.(2) The resource scheduling strategy of FIFO, computing capacity and fair distribution strategy which are commonly used in Hadoop are studied. The system throughout and system response time are compared through the simulation experiment.(3) All of the commonly used resource scheduling strategyies are not consider the tasks with the time constraints, so in this paper propose an starategy LLFD which is based to the LLF scheduling strategy. The local management algorithm, the master of global scheduling algorithm and the local system node scheduling algorithm.(4) The performance of the resource scheduling strategy LLFD on CloudSim simulation platform. Through the simulation results, it can be confirmed that under a large scale of infrastrucure environment, the strategy can be maximized the number of executed tasks, while minimized the number of nodes which can meet deadlines, then minimized the penalty rate of over deadlines with an acceptable time.