The Research And Implementation Of Metadata Cache Backup Technology Based On CEPH File System

More and more data is demanded to be stored under the driven of rapid developments of information science and technologies. According to IDC reports, the amount of produced data is expected to grow approximately 50% every year, and cloud storage technologies are born to store this massive amount of data since single node storage technologies don't fit them well. In the distributed file system of cloud storage technology, the metadata amount is very small, but the proportion of access request is more than 50%. Therefore, improving the performance of metadata accesses can effectively improve the performance of distributed file systems.Based on research and analysis of the Ceph file system, two cache backup schemes are proposed: the log cache backup scheme and the metadata cache backup scheme. The log cache backup scheme reduces metadata access delays by caching logs and reducing the time of storing logs to server clusters. The metadata cache backup scheme improves metadata access performance by buffering metadata information, eliminating the time of recording and storing logs, reducing metadata updating to data service clusters. In order to prevent the loss of cached data in metadata servers, a cache data backup scheme is proposed. Compared to the metadata management subsystem of the Ceph, the log cache backup scheme and the metadata cache backup scheme can effectively improve the access performance of metadata. Compared to the log cache backup scheme, the metadata cache backup scheme enhances the metadata access performance more effectively. However, the amount of metadata cache is more than that of the log cache, and the amount of data that needs to be back up is more than that of the amount of data, which takes up more space for the cache memory.Finally, based on open source codes of the Ceph file system, two cache backup solutions are implemented, and performance experiments of two proposed cache backup schemes and the Ceph file system metadata management subsystem are under the vdbench and the filebench test tools. Compared to the performance of the Ceph metadata management subsystem, Experiment results show that performance improvements of the log cache backup scheme are up to 11.5%, and performance improvements of the metadata cache backup scheme are up to 15.7%. So these two proposed schemes effectively improve the metadata access performance of the Ceph file system.