论文部分内容阅读
工作站机群系统需要有一个高性能的并行文件系统以适应高速输入、输出数据处理的要求，而提高并行文件系统性能的关键是合作缓存技术．文中提出了一种基于结点的分布式合作缓存管理算法 Ｄ Ｃ Ｃ．该算法综合了基于全局管理器和基于局部信息两种算法的优点，给出了一种位于结点机上的主块信息站的方法，通过对主块信息站的信息进行维护来达到对全局信息的跟踪．算法采用积极的局部信息维护策略，提高了主块位置判断的准确性．与 Ｇ Ｍ Ｓ算法、 Ｈｉｎｔｓ 算法相比， Ｄ Ｃ Ｃ算法的开销较小且具有较好的可扩展性
Workstation cluster system needs to have a high-performance parallel file system to meet the high-speed input and output data processing requirements, and to improve the performance of parallel file system key is the co-cache technology. In this paper, a node-based distributed cooperative cache management algorithm is proposed. The algorithm combines the advantages of two algorithms based on the global manager and the local information, and presents a method of the main block information station located on the node machine. Through the maintenance of the information of the main block information station, the global information Tracking. The algorithm adopts an active local information maintenance strategy to improve the accuracy of the judgment of the position of the main block. Compared with the GMS algorithm and the Hints algorithm, the D C C algorithm has less overhead and better scalability