| Summary: | 碩士 === 中華大學 === 資訊工程學系碩士班 === 88 === Most of researches in load balancing are based on Unix systems. Unlike Unix system, it is difficult to collect the workload index on Win32 platforms because of Microsoft do not release too much information about the system resource on Windows NT, 95 / 98 platforms. In the thesis, we provide a mechanism to find the system workload index on Windows NT and Windows 95 / 98 operation systems by querying their registry so that we can decide the workload level of them. By using the technique, a distributed computing environment combined Unix and Win32 platform is constructed (Java Load Balancing System-JLBS). Also, we apply neural based fuzzy logic control in our dynamic load balancing algorithm to enhance the performance of JLBS. The scheme can adjust the fuzzy membership function dynamically according to the feedback information. Therefore, our proposed load balancing algorithm has learning capability. The experimental results show that our proposed algorithm has better performance comparing with other classical load balancing algorithms.
|
|---|
