In this paper, we study an adaptive load balancing algorithm in the homogeneous distributed systems in which only the local status information is used. The parameters affecting the performance of the load balancing algorithm are investigated. In order to analyze the effects of service disciplines of the processor, we study two service disciplines: first come first serve and priority queueing. All tasks in the system are treated the same in the former discipline. The latter discipline divides tasks into two classes, local tasks and remote tasks, and assigns different priorities to them. Five priority queueing disciplines are compared. We use the Z‐transform and the Matrix‐Geometric solution techniques in the mathematical queueing analysis. The simulation results and the numerical results are presented and used to shed light on the characteristics of the load balancing process. Finally, we compare the results from these two approaches and investigate the accuracy of the analytic modeling.