In multi-server environment there are many servers which are able to process job requests. So we bave to design a mechanism that selects appropriate servers for processing each job request while maximizing server throughput and minimizing average response time of requests. Conventional methods ac~ opt the load of each server as criteria of server selection. that is, they select a server whose load is not bigger than the others. In this work we propose an approach that uses the degree of server performance, server load and the estimated service time of requested job as guidelines of server selection. We incorporate fuzzification techniques and expert knowledge in this approach. Comparing the performances c~f our approach to that of conventional one, experiments show that the proposed approach provides better performances.