Abstract: A distributed Web log mining system model (DWLMS) is presented. Based on the analysis on the procedure and algorithm of Web frequent access pattern mining, the more general incremental updating algorithms of local frequent paths (LFP) and global frequent paths (GFP) in a distributed database system based on DWLMS are proposed for discovering the frequent access paths. Some troubles produced by real time incremental distributed Web access information and more communication data are solved better by the algorithms. The method was realized simply and tested with real world Web log information in laboratory, and the results show that the algorithms are valid.