The dataprocessing center of the meteorological satellite ground application system manages a huge amount of data each day， so high timeefficiency and reliability is required.In view of the high requirement and load balance in the cluster system， in use of the experiences of the Ants algorithm， Reserving Resource， Dynamic Priority strategy, and virtual server application concepts， a parallelscheduling algorithm (a dynamic pheromone weighting algorithm based on performance forecast and link weight calculation) is presented, which can be applied in the dynamical load balance of HPC, help the cluster find the optimum server to deal with the requirements of application rapidly and exactly， thus realize the high dateprocessing capability of the data processing center.