Design and Implementation of Error Rollback Mechanism for MapReduce

• Main Authors: Li-Yu Lu, 盧力瑀
• Other Authors: 張晉源
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/06903390387172592623

碩士 === 樹德科技大學 === 資訊工程系碩士班 === 101 === With the vigorous development of the Internet, the Internet access has become part of people’s lives. Therefore, it is important for the using huge resources on the internet effectively. The traditional computing is not applicable to the great amount of data nowadays. Consequently, accomplishing the data by using the discrete-computing architecture has come to be the trend. This research confers the platform of the open source software - Apache Hadoop which is using the parallel-computing architecture. When the default architecture, Hadoop MapReduce , comes about the condition of JobTracker Fail ,it will cause the computing disconnection. Besides, when the JobTracker rollback, the Hadoop can’t return the place which affected computing disconnection last time. Therefore, this research brings up the framework which is using the Queue and the Memcache. Because of that, when we face the JobTracker Fail, we can still return to the original state.