In the large-scale distributed environment load-balancing is a challenging task for maximizing the performance of the system. The global state of a large-scale distributed system is swiftly and dynamically changing, and it is very difficult to accurately model the system using a typical approach. In this paper thus we propose a new approach for improving the performance of distributed system using an intelligent fuzzy grouping approach. It utilizes a membership graph representing the amount of CPU time and memory space used for inferring the service priority and then load distribution. Extensive computer simulation reveals that the proposed approach allows consistently higher performance than the existing approaches in terms of response time and throughput for various numbers of servers and tasks. Also, it reveals that fine-grain membership graph and fuzzy inference rule allow higher performance than coarse-grain model.