基于Hadoop的地震属性数据存储结构

摘要

地震属性分析在储层评价、地层岩性解释等方面有着重要的作用.单机环境地震属性分析计算比较耗时,使用Hadoop分布式并行计算框架,可以缩短计算时间.单一地震属性以SEGY格式独立保存,多种地震属性上传至HDFS时被分发至不同节点,而计算时通常使用多种地震属性共同参与计算,增加了节点之间的数据传递的开销.通过改变地震属性数据组织形式,抽取同一坐标地震属性进行组合,并把坐标位置相邻的地震属性组合到一个文件中,减少集群节点间不必要的数据传递.实验表明相比于传统方法,使用MapReduce完成地震属性分析相关计算用时明显减少,并且计算用时随着集群中节点数目的增加而减少.%Seismic attribute analysis and calculation in stand-alone mode is time-consuming,while using Hadoop distributed parallel computing framework can significantly shorten the time.Single seismic attribute is saved in SEGY file independently while multiple seismic attributes are uploaded and distributed to different nodes in HDFS.Generally,a variety of seismic attribute are used in calculation lead to the cost increasing of data transfer between nodes.By changing the organizational form of seismic attribute data it extracts and conbines the seismic attributes at the same coordinate.Also combining the seismic attributes adjacent to one file can reduce the unnecessary data transfer between nodes in cluster.The experimental results show that seismic attribute analysis with MapReduce significantly reduces the time compared to the traditional methods and furthermore time costs decreases with the number of cluster nodes' increaseing.