The Research Of The Local File System For Object-Based Storage Device

Driven by the requirements for extremely high bandwidth and large capacity of the storage system, storage system architectures are undergoing fundamental changes. The object-based storage model, in which files are made up of one or more data objects stored on intelligent Object-Based Storage Devices(OSDs), is a promising new model. Object-Based Storage Device(OSD) manages its own storage space, and exports an object interface.The workload presented to the OSDs will be quite different from that of general-purpose file systems, so the effictiveness of OSD will be limited when using general-purpose file system as its storage manager. OSDFS, a local file system specially designed for OSDs, is presented at this background, in order to improve the performance of OSD storage management.This thesis focuses on the design, implementation and performance analysis of OSDFS. OSDFS adopts some design choices based on the features of OSD workload. Compared to general-purpose file systems such as Ext2, OSDFS has some unique features in four aspects. Firstly, OSDFS is a file system based on extents, compared with block based file systems, extent based file systems are more efficient for large objects access. Secondly, object names are organized in a B+ tree, which is very suitable for large directory index in the OSD flat namespace. Thirdly, free extents buckets and the preallocation policy is used to manage free space allocation in order to keep objects continuous on disk and reduce the external fragmentation.Finally, we evaluate the performance of OSDFS. The benchmarks show that OSDFS outperforms Linux Ext2 when dealing with OSD workload, and OSDFS ages slower than Ext2 during the long term object operations.