| AFS (Andrew File System) is a distributed file system that enables usersto share and access all of the files stored in a network of computers. AFSis based on client-server architecture for sharing files and providingindependent allocated transparent environment for data.Based on the research of architecture, features and relationships amongmajor services in AFS file system, the thesis analyses defects found in currentsystem resource management method, then proposes and designs intelligentmonitor service to optimum system performance automatically. Intelligentmonitor service is consisted with three main parts, first is the improvementof resource management, second is the control of the resource abanudancy, andthird is the feedback of system error conditions. The most fundamentalcomponent in service design is to establish and update the strategy in systemresource management and to obtain effective management strategy throughreasonable clustering and priority sorting in network servers. The Max-mindistance clustering algorithm and typical load balance algorithms arereferred and combined in resource management strategy with some improvement.All components in the design and realization of the intelligent monitorservice will be illuminated in the thesis. Two components, server clusteringand copy order adjusting, which are related with the establishment andadjustment of resource management strategy will be discussed with emphasis.By designing different scenario for experimenting and comparing the testresults with different resource management strategies, intelligent monitorservice has been proved to be reasonable in design and high-efficiency inpractical applications.The research achievements in the thesis indicate significant value in therealization of optimum management and load balance in distrbuted file systems. |
