A Study On Server-side Cache For Object-Based Distributed Storage System

Nowadays with the rapid development of the information technology, storage technology and cloud technology, a number of data sets are generated everyday. More and more storage systems have raised up to peta-scale. The distributed file systems are becoming more mature while they are used widely as the user-level of mass storage system. They are able to connect shared folders with the file servers in transparent ways. They map the data which are from different locations into the single logical level to make themselves more convenient. People have been concerning about their high efficiency, generality, reliability and expansibility. On the one hand, Object-based Storage System (OSS) are able to provide higher performance than some other distributed storage systems, which attracts rising attention in the storage field. However, the I/O rate of OSS is effected by that there is no cache to provide high rate aiming at hot files in heterogeneous storage environment. So it’s necessary to research on the high efficiency of the distributed file system in the field. On the other hand, Web caching can improve the performance of the storage system, reduce traffic congestion and increase satisfaction levels of clients using the high-performance devices to storage the data. As a result, it’s significant to research on the web cache based on OSS.After reading the domestic and international papers, this paper discusses the key technology about the OSS and web cache. To solve the problem that there is no cache to provide high performance for popular objects in heterogeneous storage environment, we design and implement a server-side cache upon object-based distributed file system(Server-side Cache Scheme for Object-Based Distributed Storage System, OBSC). The work of this paper is organized as follows:(1) This paper discusses some domestic and international storage platforms, and analyzes the storage technology of object-based distributed file system. Based on the problem in heterogeneous storage environment, this paper mend the traditional data accessing pattern and propose a cache scheme based on OSS. Based on the framework of the server-side cache, we design a operating protocol which control the communication of the clients, metadata servers and OSD cluster. At last we analyze the consistency and reliability of this scheme.(2) According to the cache scheme based on OSS this paper implement a server-side cache based on object-based distributed file system(OBSC). OBSC optimize the mechanism of the metadata servers and OSD cluster in order to cache the files which are accessed frequently. (3) According to the server-side cache scheme mentioned above, we discuss the traditional replacement policy and the accessing problems of the cache. We propose frequency based replacement strategy which considers the characters of object-based distributed file system and the placement of the data to improve the I/O rate of the overall system.(4) According to the research above, we present the overall design and implement of the server-side cache for object-based storage system. Also, we take some experiments to demonstrate its efficiency and feasibility in terms of the performance test and overhead test.