Design And Implementation On Memory Management Module For Large-scale Seismic Data

With the rapid development of computer technology in the field of geographicalexploration, deep exploration will produce a ultra-large-scale seismic data SEGY file, the filesize can reach Gigabyte or Terabyte order of magnitude. Traditional earth science applicationsused in the data processing load, the repeated use of the mode. However, when the size of datafar exceeding the memory traditional loading mode will not be able to used for dataprocessing. To this end, the seismic data file format and memory management two means tosupport the ultra-large-scale data. This ultra-large-scale seismic data processing means themonopoly of foreign technology.The author of this article where the research team VSG company made reference toforeign commercial software Open Inventor based on octree structure created a large-scaledata-oriented file organization system LDM. The memory management module used ingeographic Seismic ultra-large-scale geochemical data LDM show.This paper designed anddeveloped a memory management module, including buffer management and prioritymanagement component assembly of two parts: the buffer cache and memory managementcomponent is responsible for the management of the data space, as well as a reasonableallocation of memory resources; priority management component is responsible for thememory most likely to be retained next time accessed data blocks and supports priority pluginscalable functionality and user-defined priority algorithm import. Buffer managementcomponents can be loaded into memory on multiple files simultaneously seismic datamanagement supporting dynamic file buffer size. Priority management component supportsmultiple priority algorithm to work together and take advantage of multi-threading means toachieve single file priority management and rapid response. Priority management componentprovides users with several classic pre-priority algorithm, including the first-in, first-outalgorithm, the least recently used algorithm and other typical algorithms. At the last part ofthe paper, the visualization module test example, to support20times the size of the datareaches the memory size of the seismic data load, display, and supports multi-resolutionreal-time switching.The modules are in different memory size, number of visits, randomconditions, file access frequency blocks to create a set of test conditions to hit rate benchmarks designed to simulate experiments. Experimental results show that the first-in,first-out algorithm and the least recently used algorithm to good effect, and memorymanagement module in the least recently used algorithm can achieve3%to30%of theaccelerating effect.