Design of PPU framework for processing ordered data blocks in the cluster environment

Abstract

With the development of parallel computing, processing mass data in the cluster environment becomes a hotspot in research. A typical scenario is that the data blocks needed to be parallelization processed is ordered, but there is no common framework adapting to this scenario. Under this background, a common framework named PPU is proposed in this paper to resolve the parallelization of processing ordered data blocks. With the preemption between processes, the PPU framework can ensure the efficiency in parallel data processing. And the mapping table is used to keep the data blocks in the correct order after parallel processed. Experiment results show that the PPU framework for processing ordered data blocks is feasible.