Fast Loading Mechanism Of Large Seismic Data Based On Octree Structure

With the development of hardware technology, the magnitude of theexperimental data in the field of deep exploration grows rapidly. Although theperformance of the computers used for processing data continues to improve, theexplosive growth of data scale is still beyond the development of memory. In thepractical Geoscience activities, the scale of seismic data tends to reach dozens of GBor even TB level, while it is hard to completely load the data into memory invisualization process. Traditional volume data rendering process completely load datainto memory in one time. Rendering module can only get data from memory.Obviously this way failed when the scale of data exceeds the memory size. Therefore,achieving the management of large data in storage and loading process, establishingsuitable logical data organization and fast loading mechanism for the fast retrieval andvisit of large volume data is important in the research work for large data.Based on the study of the octree structure used for large data storage, this paperdesign and implement an octree file format and file index table firstly. By storage largedata in octree structure, the natural multi-resolution mechanism of octree is obtained.Real-time rendering is guaranteed by sacrificing the resolution.In order to meet the demand for fast memory loading of large data duringapplication, this paper propose a multi-file, multi-resolution, multi-threaded,multi-loading priority loading mechanism and gives a top level architecture design. Forthe fast data accessing in the architecture, we design and implement a scheme whichconvert the original loading request in two stages and assigned the result to theloading thread eventually. Via the first conversion by the synchronous orasynchronous data accessing interface oriented the user needs, the user’s originalrequest is interpreted as a list of data blocks and passed to the request accessingmanagement module. Through the further broken down of loading request, the resultis assigned to the threads in order to achieve the load balancing in a multi-threadedloading process.