面向地震数据交互分析场景的高效分布式缓存框架

随着高精度地震勘探的发展与应用,地震数据规模在飞速增长,大数据分布式存储系统(如Hadoop分布式文件系统(HDFS))虽然可以存储海量地震勘探数据,但面对单炮记录超过万道的大规模数据时,数据实时访问时间延迟较长,从而导致交互操作卡顿,影响操作流畅性,降低了工作效率。对比分析了在HDFS以及分布式内存存储系统Alluxio上地震数据的读取性能差异,利用Alluxio分布式内存的性能优势,在HDFS底层存储基础上,设计了一种面向地震数据交互分析场景的高效分布式缓存框架,可以有效降低大规模地震数据实时访问的时间延迟,进而提高交互分析工作效率。实际数据验证结果表明,该分布式缓存框架在大规模地震数据交互分析场景下,数据访问性能提升30%以上,具备实时性能优势,能够满足交互分析需求。

With the development and application of high-precision seismic exploration,the scale of seismic data has been growing rapidly.Although big data-distributed storage systems,such as the Hadoop Distributed File System (HDFS),can store massive amounts of seismic data,the real-time data access delay is relatively high when a single shot records more than 10000 channels of seismic data.The delay causes an interactive lag,thereby affecting the fluency and efficiency of the operation.In this study,the differences in the reading performance of seismic data were compared and analyzed on HDFS and the distributed memory storage system Alluxio.Based on the performance advantages of Alluxio and the underlying storage of HDFS,an efficient distributed cache framework for interactive analysis of seismic data was designed.The framework can effectively reduce the real-time access delay of large-scale seismic data and improve the efficiency of interactive analysis.Practical application in large-scale seismic data interaction analysis verified that the proposed framework can improve data access performance by more than 30%,which has real-time performance advantages and can meet the requirements of interaction analysis.