Distributed Parallel Cache Management Mechanism And User Consistency

With the development of the society, data in all kind of fields increase in an amazing speed, especially in the improvement of the Internet Age. Meanwhile, huge amount of information needs to be stored in the digital way. As a result, how to store and manage the digital information is becoming the focus concerned. In the 70's and 80's, these data were stored in the expensive middle or mini servers and system administrators had to back up the data periodically, which needed huge amount of investment. However, companies still took the risk of suffering from data lost, if these servers get fatal errors. With the deeper research of distributed and parallel system, with the high development of hardware, with the high-speed network will come to be mature, and with the operating system of Linux has developed rapidly, there were many distributed server system based on Linux to store and manage data developed. These systems are generally composed of high-performance PC servers interconnected through the high-speed LAN. Servers, which managed by distributed and parallel operating system, can work independently or cooperate with each other to store data, which can enhance the reliability of the data and improve system's performance compared with single server.DPFS (Distributed and Parallel File System) discussed in this thesis is the core of distribute and parallel system. Based on Linux, DPFS is designed specifically for the distributed and parallel servers. It is used to manage data stored in the system intelligently. And the high reliability of the system can be achieved. DPFS also improved system's ability of service.The thesis will describe the overview of the development of the data storage and the new demand of it. It also introduces the effect of the DPFS. And the problems must be considered in the design of the DPFS. The paper analyzes requirement we must meet when designing DPFS and presents the logic structure of DPFS. The mechanism of cache management and the users consistency of operating system are discussed in detail. The dynamic probe and feedback algorithm of cache sending is described particularly.