Speaking on the Landmark booth to a good audience, Chuck Mosher (ConocoPhillips) traced the history of the JavaSeis project*. ConocoPhillips, Arco and Chevron all contributed intellectual property to the project which is now in the public domain. The main value proposition of JavaSeis is its support for parallel input/output (I/O) and the ‘parallel distributed array’ concept.
Filing cabinet
Typical seismic processing is done on a trace, or on an ‘ensemble’ of traces, at a time. 3D ‘process-lets’ produce small but cascading quality improvements. Unfortunately these are hard to implement and manage on clusters, which is where JavaSeis comes in. JavaSeis’ ‘filing cabinet’ metaphor allows for massively parallel I/O and introduces ‘true’ multi dimensional arrays à la Fortran into Java. Code is written for serial processes or for a single processor and JavaSeis manages parallelization behind the scenes. Unlike many conventional approaches, data is processed ‘in situ’ and not moved around.
SeisSpace
JavaSeis is the default storage format for Landmark’s SeisSpace seismic processing environment. SeisSpace also leverages JavaSeis as the native I/O subsystem for large scale parallel data processing. The latest release of SeisSpace now supports parallel processing. And significant performance of sorting and FXY deconvolution has been reported. It is now possible to work with an ‘ensemble’ and perform true 3D operations with 3D algorithms. A new 3D visualization toolkit has been built for processing as opposed to interpretation visualization tools.
* http://javaseis.sourceforge.net.
© Oil IT Journal - all rights reserved.