Research And Implementation Of Log Management System For Distributed Software

With the rapid development of internet and the increase of network information, the distributed application system based on very large data scales up increasingly. Consequently, the parts of distributed application system become more gigantic and complex, so it brings troubles to the debugging of system and knowing the status of system. It is a research work of general and practical significance to solve the problem of using log information generated by distributed application system effectively for analyzing system status and enhancing the abilities of pre-alarm to satisfy its urgent demands of performance, robustness and scalability.This thesis studies the issue of log management for distributed application system based on StarTPMonitor. Then, on the base of our study, we develop a log management system for analyzing status of distributed system. The primary contributions of this thesis include:1. First, we analyze the application background of distributed log service and the development trend of log technology in different area. At the same time, we describe deficiencies in some middleware products of Log Service on satisfying the requirement of current system.2. Second, we introduce the whole struct of StarTPMonitor, include where Log Service is in it. Then, we simply describe the characteristics of two master services which will use Log Service, and describe the requirements on Log Service.3. Based on the analyses of the requirements on Log Service, we design a Log Management System for Distributed Software - LMS (Log Management System, LMS). We discuss the architecture, log format, log management patterns, means of get alarm information of LMS detailedly, include using "Two Buffers policy" to improve transfer speed of log files, using improved clock synchronization algorithm to solve the problem of ordering log events from multi-node, and automatically generate alarm rules by mining association rules to enhance the efficiency of log analyzing.4. We implement LMS based on StarBus. The three-segment architecture of LMS implements fastly recording. It provides dynamic object management for system debugging in running mode and level control of outputing log datas. It implements file transferring in a transparent way to application using CORBA any type to code and encode log files. Also, it uses XML technology to store configuration datas to ensure the flexibility of LMS.5. Finally, the experiments on the functionality and the performance have also been carried out. Empirical results demonstrate that LMS works correctly, and the overhead incurred is very small, and LMS can satisfy the performance, scalability and availability which StarTPMonitor requires on.