Achieving good load balance in cluster environment is difficult due to its dynamic nature. Besides, the data to be processed may be transfered from geographically distant sites at a fluctuating transfer rate. This paper proposes a load balancing algorithm to deal with such fluctuating transfer rate. The novelty of the algorithm is to use buddy concept in processing node grouping and the data distribution statistics obtained at the source sites. We use single join processing in database query processing as its application.