Researches About Cloud Computing And Expolit And Test Hadoop Program

As we have reached the extreme of computer's physical abilities, Moore's Law does not work any more. This directly pushes researchers to think many times in order to improve computer's operational performance. On the basis of this, a dual-core machine and a quad-core machine come into being. In fact, this attempt is practical and successful.Thanks to that many techniques, such as parallel computing, internet and virtualization become more and more mature with pursuing severs in high ability, high storage, high reliability, high extendibility, people come up with the idea, called cloud computing. In 2007, as the pioneer, Google came up with the concept named cloud computing firstly. The method to realize it is to combine several cheap computers together and make them work cooperatively, which is equal to or even better an expensive super sever. In the same year, IBM, Microsoft, Yahoo, and HP entered the development of cloud computing sequentially. Nowadays, all of IT giants regard cloud computing as one of main strategies in the future development. In addition, all of commercial media hold positive attitude to the development of cloud computing. Thus, there is no doubt that could computing will have bring commercial preview and practical needs.This paper mainly introduce the generation, development current situation of cloud computing; compare internet computing with cloud computing; analyze parallel computing language MPI and Erlang. Besides that, this paper also includes the overall framework of open source cloud computing platform MapReduce. This paper mainly covers two parts of Hadoop-MapReduce and HDFS, where analyzes the source code and procedure of MapReduce, how to create operational environment and situation for Hadoop and to develop management software about Hadoop cloud computing platform. In addition, the paper also introduces how to program the framework of MapReduce.This paper focuses on:1. Analyze the source code of MapReduce and the call between JobTracker and TaskTracker during MapReduce.2. Design a program to arrange Hadoop, whose main functions are open/close Hadoop, demonstrate node information of Hadoop and show the changes in environment and so on.3. Study and design a new scheduling algorithm, the scheduling algorithm is different from the traditional FIFO, the polling time slice, priority mechanisms. Its feature is more flexible than all the others. Believe that we can to all of the relevant personnel scheduling a new way of thinking.This paper tries to analyze how MapReduce operate and its procedure in depth. At the same time, can lead a programmer who does not understand Hadoop can program MapReduce smoothly on this platform.