Ext2 file system is optimized for the management of small files, however, the performance of managing large files is not perfect. The Object-based Storage Device of HOSS[1] (Huge Object-Based Storage System, HOSS) adopts ext2 as the underlying file system, so the performance of managing large objects is limited. Two strategies to improve the reading performance of large objects in OSDs are proposed in this paper. Firstly, local relativity of data blocks on disk is utilized to reconstruct the data block index addresses as Successive Data Blocks Index Array (SDBIA). SDBIA will be taken as one attribute of the object and then the performance of getting the data block index address of large objects will be improved obviously. Secondly, successive blocks allocation strategy is adopted. The experiment shows that these two strategies are effective for the performance of reading large objects.
展开▼