Research Of Seismic Data File System

In recent years,thanks to the human society demand for petroleum,petroleum exploration technology is in rapid development.Seismic exploration is the main method for petroleum exploration,with the continuous innovation of seismic exploration technique in seismic data acquisition,data processing and data interpretation,seismic exploration data is in massive growth.At present,people have developed high-performance computing against the vast amounts of data computing,it makes the speed of seismic data processing become faster and faster.But the data IO bottleneck in seismic data processing has seriously hindered the speed of seismic data processing.Therefore,how to improve the reading speed of seismic data is a problem that shall be solved immediately.In view of the seismic data IO bottleneck problem,a distributed file system is used to become the storage system.The characteristics of each distributed file system are not the same.For example,HDFS is suitable for large files,and it has optimization for the large file reading,and Fast DFS is suitable for small files.The characteristic of seismic data reading is that the data you want to get will be small in size but large in amount.The current distributed systems are not in conformity with the characteristic of the reading mode for seismic data.Thus the IO bottleneck problems in seismic data reading still exist.Based on the research of the characteristics of the main seismic data format SEGY and general processing of seismic data,this thesis analyses the cause of the IO bottleneck when the seismic data is processed.And then,a seismic data storage system which is based on Fast DFS will be realized,it is called SDFS(Seismic Data File System).This system puts forward the solution for IO bottleneck in data processing.The solution as follows:first,the seismic data files will be partitioned by the trace size,it raise the efficiency of data reading;Second,to partition file macro block to block and the using of parallel reading technology in disks improve the speed of seismic data reading;Finally it also puts forward the study of the reliability in storage system,namely using a full backup to increase system reliability.In order to prove the effectiveness of the proposed scheme,this thesis realizes the SDFS system and the main functions of the system have been designed.And then,two groups of the experiment have been carried out to test the performance of the seismic data reading.The first group of experiment tests the effectiveness of the parallel reading technology in disks,the results of the experiment show that,when the number of hard disks in each node is increasing,the performance of system increases,the speed of reading is almost proportional to the number of disk;When the data block size increases,it is advantageous to sequential reading,and when the data block size decreases,it is advantageous to random reading.The second group of experiment is a test for the system performance by contrast experiment,the contrast system is Fast DFS.The results show that the performance of sequential reading and random reading is better than Fast DFS's.Integrating the results,a conclusion can been put forward.That is the SDFS system can help improve the IO performance of seismic data reading.