论文部分内容阅读
现在三维地震勘探数据通常都在TB量级,地震勘探处理中文件I/O的时间花费越来越引起了人们的关注。常规的偏移成像算法分块通常按照先inline再xline的顺序来分块,这样分块看上去比较简单直接,而且算法实现上比较容易。但这样分块会导致整个偏移过程中总的文件I/O量与原输入数据相比特别大。常规分块算法对小规模地震数据文件处理时,增加的文件I/O所花费的时间与偏移的时间相比不是很明显,因此因分块导致的文件I/O量的增加常常被人们所忽视。当处理的地震数据量大时,如现在的TB量级的三维地震数据,常规的分块算法导致I/O量的增加就突出了,通常达到了原数据的700~800倍,相当于读入700~800TB的数据,由此所消耗的读取时间也是惊人的,若换算到由单结点去读取,则需要花费4个月的时间。为此,我们提出二维分块算法,其基本思想来源于偏移成像的有效地震道分布依赖于定义的偏移孔径,将成像分块的方式从常规的inline优先改为inline和xline方向同等,形成一个个接近正方形的成像区域。在结点能装载的成像区域同等情况下,二维分块算法完成成像所需要读入的地震道数量比常规分块算法要明显的少。
Now that 3D seismic data is usually on the order of magnitude of TB, the time spent on file I / O in seismic exploration is drawing more and more attention. Conventional offset imaging algorithm block usually follow the first inline xline order to block, so the block looks relatively simple and direct, and the algorithm is easier to implement. However, such a chunk causes the total file I / O volume to be particularly large compared to the original input data throughout the offset. Conventional Partitioning Algorithms When dealing with small-scale seismic data files, the increased file I / O does not take much longer than offset, so the increase in the amount of file I / O due to blocking is often seen by people Neglected. When dealing with a large amount of seismic data, such as the current three-dimensional terabytes of terabytes of data, conventional block-based algorithms result in an increase in the amount of I / O, typically 700 to 800 times the original data, which is equivalent to reading Into the 700 ~ 800TB data, which consumes the reading time is amazing, if converted to read from a single node, you need to spend 4 months time. To this end, we propose a two-dimensional block algorithm whose basic idea stems from the fact that the effective distribution of the seismic traces of the migration imaging depends on the defined offset aperture, changing the way the image is segmented from the regular inline priority to the inline and xline directions , Forming a nearly square imaging area. When the node can load the same imaging area, the number of seismic traces to be read by two-dimensional block algorithm is much less than that of the conventional algorithm.