AbstractTo a large extent, the load balancing algorithm affects the clustering performance of the computer. This paper illustrated the common load balancing algorithms and elaborated on the advantages and drawbacks of such algorithms. In addition, this paper provides a kind of balancing algorithm generated on the basis of the load prediction. Due to the dynamic exponential smoothing model, such an algorithm helps obtain the corresponding smoothing coefficient with the server node load time series of current phrase and allows researchers to make prediction with the load value at the next moment of this node. Subsequently, the dispatcher makes the scheduling with the serve request of users according to the load predicted value. OPNET Internet simulated software is applied to the test, and we may conclude from the results that the application of such an algorithm acquires a higher load balancing efficiency and better load balancing effect.