Abstract: Middlcware in Internet cnvironmcnt,typically Wcb Application Server(WAS)，needs to provide the performance guarantee and optimization services effectively for Internet applications. Thread pooling technique is a common method of performance optimization. With the consideration of the features of Internet applications,the management of thread pool in middleware needs to adjust dynamically on the basis of perceiving the context at run-time. However,how to find out effective influencing factors which make the adjusting results having better adaptability remains to be discussed further. The paper firstly presented a dynamic thread pool model of application server based on M/M/1/K/二/FCFS queuing system;then,the paper studied a mechanism which imports heuristic factors and rules for reflecting the context at run-time. It can realize dynamic adjustment of thread pool size more effectively so that thread pool size is well adapted to resources change. The experiments verify the effective influence of heuristic factor that exert on a山ustment of thread pool system.size and show that the presented management mechanism can significantly improve the performance of system.

Key words: Thread pool, Middleware, Queuing system, Heuristic