With the development of hardware technology of computer architecture and the increasing requirements of mass data storage, new storage devices and advanced storage architectures are needed to bring high-performance mass storage.Solid State Disk won its edge over hard disk in storage systems, for it provides better I/O performance. At the mean time, Object-based storage architecture beat the other storage architectures. In this paper, we investigate SSD and Object-based storage, and propose an Object-based storage acceleration system with SSD.What we have done are as follows.(1) We study solid state disk techniques, architectures and accessing features. Then we analysis the different I/O characteristics of hard disk, Flash-based solid state disk and DRAM-based solid state disk respectively.(2) We study the development of storage architectures and focus on Object-based storage architecture and Object-based storage file systems.(3) Based on Lustre file system and the characteristic of solid state disk, we propose an Object-based storage acceleration system with SSD.(4) We analysis the I/O access patterns and multi-experts system, as well as classic algorithms of cache replacement and prefetching. Then we design the algorithm of multi-experts object management.(5) We implement the prototype for the algorithm of object cache management, along with replacement expert modules as FIFO,LRU,LFU and MRU. The result of test under different I/O access patterns demonstrates the adaptivity of our algorithm.(6) We implement the prototype for the model of hybrid OST and make a comparison of latencies under random I/O access pattern respectively.Some of the key techniques and design ideas of our work are also helpful for the research of other storage systems. |