A case study on predictive method of task allocation in stream-based computing

Static methods of task allocation have less overhead at run-time but they are difficult to adapt to the environment. We propose the predictive method in order to increase the adaptability of static task allocation methods. In this method, optimal allocations are decided by using system dependent factors which have been measured. In addition, we introduce stream-based computing which is suitable for data streaming as a HPC method. In this computing model, it is easy to predict computing costs. In this paper, we describe the evaluations of our predictive method in the case of applying it to stream based computing