Testing And Evaluating Technology Research Of The Distributed File System On The Cloud Platform

With the continuous development of cloud computing technology and itspopularization, the concept of cloud storage also arises at the historic moment, inwhich the distributed file system is the core foundation, responsible for the datastorage. Also, with the improvement of social informatization degree, thedependence on the information and data is on the rise, which means the reliability ofthe data is more and more important. Fault tolerance is an important standard tomeasure a system reliability. The stronger the tolerance of the system is, the morereliable service the system can provide. Therefore, the evaluation of thefault-tolerant ability in the distributed file system on cloud computing platform hasimportant research significance.At present, most of the researches at home and abroad of distributed file systemon the cloud platform are focused on performance testing and optimization. Severalinvolve in the fault-tolerant ability evaluation, but their means of testing is relativelysimple. Therefore, it is necessary to conduct in-depth research on the evaluation ofthe fault-tolerant ability in the distributed file system on the cloud platform.Based on the analysis of HDFS and TFS （two typical cloud distributed filesystems）, which includes researches on their system structures, input and outputprogresses and inner fault tolerance mechanisms, by comparing their similarities, aset of fault-tolerant ability evaluating methods is presented. According to the ideaof fault injection, a set of fault injection tool sets is designed and implemented,respectively, for the nodes and process of the cloud distributed file system, files anddirectories, data verification mechanism, internal interconnection network faultinjection, to simulate various types of faults that may occur in the application.In order to verify the feasibility of the fault-tolerant ability evaluating methodsfor the distributed file system on the cloud platform, a small HDFS distributed filesystem is built. In the meantime, HDFS is taken as an example to product theassociated fault injection experiments. Through observing phenomena and analyzethe anomalies recorded in the log file, the fault tolerance ability of the HDFS can beevaluated and validated. Moreover, methods for improvements will be proposed.