Research And Performance Testing Of The Ceph Distributed File System

As the storage subsystem of Distributed Systems, Distributed File System thatcan effectively solve the I/O bottleneck problem has become a research hotspot in thecorresponding industry. Ceph is an emerging Distributed File System whose firstversion was released in June2012. All aspects of this Distributed File System, such asarchitectural and structural features, performance, availability and expansibility and soon, is in badly need of testing, verifying and studying.The research object of this paper is Ceph Distributed File System. F irstly, wedraw up a detailed analysis of two key technologies of the system’s architecture andarchetype: CRUSH algorithm and RADOS objective storage service. Their impact onthe performance and expansibility of the Ceph system is mainly studied. In the next,we set up a test environment design system test cases according to the results ofprevious research and then use the cases to conduct a comprehensive test on theperformance, availability and expansibility of the system. Finally, a new Journalstorage solution and a new cluster network topology are proposed on the basis of theresearch. To verify the correctness of the optimization, a further test of the optimizedsystem is conducted.The test results indicate that Ceph Distributed File System possesses highavailability high expansibility and excellent performance. Also, the two clusterdeployment optimizations raised in this paper do improve the performance of thesystem.