The framework of parallel algorithm can significantly promote the efficiency of segmented seismic data processing and reduce the processing cycle with the help of multinode and multithread parallel computation technology. This paper presents many real data tests for comparisons, and analyzes various factors accordingly which affect the computation efficiency of parallel algorithm. The comparisons tell that the bigger the computation workload in a parallel processing job is, the higher the processing efficiency is; but the processing efficiency of parallel computation cannot increase linearly. With the increase of processing jobs, the acceleration of computation efficiency will slow down gradually and an inflexion will occur at the optimum degree of parallelism. So the best degree of parallelism in a parallel framework will play a key role to keep higher efficiency in seismic data processing.%并行计算框架系统将地震数据可分割处理的特点与现代计算机群多节点、多线程的并行运算能力巧妙结合,在提升常规地震处理运算效率、缩短地震资料处理周期方面发挥了显著作用.本文通过大量数据的测试,对影响并行计算框架运算效率的各种因素进行了分析.并行计算框架对地震数据处理运行效率的提高并不是随着并行任务数线性增长,与机群I/O效率及并行处理作业的计算量有关.随着并行任务数的增加,效率提高速度会逐步放缓,当达到最佳并行度时会出现拐点.另外,并行计算框架效率的提高与数据分割粒度有关,粗粒度分割更有利.因此应用并行框架系统时,应根据机群I/O效率与并行处理作业的计算量,选择合理的并行处理任务数.
展开▼
