Performance Test And Optimization Of Distributed File System

The rapid development of the Internet not only brought great changes to our life,but also brought about the rapid growth of data volume.The amount of data people need to handle has grown from TB to PB,EB,and even ZB.The Distributed file system has the characteristics of high concurrency,high scalability,high performance,higher availability and large capacity,which makes it an ideal choice for storing massive data.Glusterfs is an open source Distributed file system,which is the main research object of this paper.Firstly,the architecture and key technology of glusterfs,especially its unique elastic hashing algorithm,are analyzed in detail.According to the analysis result design test case,build test environment,test the system performance.Includes the sequential read/write performance of the four basic volumes of the glusterfs in native,NFS,CIFS three usage modes,aggregate performance when the distributed volume uses libgfapi,and glusterfs metadata performance.Then the test results are compared and analyzed.In the test,we found that the system performs well when storing large files,especially its aggregation performance.However,its metadata operation performance is not satisfactory.To this end,we have improved the Glusterfs metadata storage method,using LevelDB to store the extended properties and metadata of each directory and file on the brick of Glusterfs.When Glusterfs performs an operation that requires changing metadata,the LevelDB is updated at the same time,and is read directly from Leveldb when Glusterfs performs a metadata query operation.Leverage Leveldb’s high performance to speed query access to metadata.Finally,the simulation experiment is carried out to verify the effectiveness of the design.The experimental results show that the metadata storage method designed in this paper can improve the data access speed of the system to some extent.