1.13 Parallel Processing in GeoRaster

There are two types of parallel processing with GeoRaster.

• Parallel execution of SQL statements

• Parallelized GeoRaster procedures

Parallel execution of SQL statements allows most SQL statements, both query and DML, to run in parallel. When a SQL statement is executed, it is decomposed into individual steps or row-sources, which are identified as separate lines in an execution plan.

All GeoRaster read-only functions such as metadata-related query operations (that is, all GeoRaster metadata get functions and SDO_GEOR.validateGeoRaster) and all single-raster cell queries (SDO_GEOR.getCellValue and SDO_GEOR.evaluateDouble) are enabled for parallel query. This means that in a multi-CPU environment, if these functions are used to query many GeoRaster objects in one or more GeoRaster tables and if the SQL statement is made to run in parallel, the GeoRaster rows are automatically divided into multiple subsets, and multiple Oracle server processes will work simultaneously to process each subset to reduce the overall response time. By dividing the work to run a GeoRaster SQL statement among multiple processes, you can more quickly maintain spatial indexes and find GeoRaster objects based on their locations, various metadata, and attributes. You can also use the pipelined and parallel table function to implement more sophisticated procedures, including parallelizing some operations on a single GeoRaster object.

Parallelized GeoRaster procedures let you specify multiple subprocesses for simultaneous processing of a GeoRaster object. Some individual raster and image processing procedures are specifically implemented to support this type of parallelism. With these procedures, you simply specify an integer number for the degree of parallelism (DOP) as an input parameter, to cause the operation to be split into that number of subprocesses to process the subsets of a single GeoRaster object simultaneously. Each of those subprocesses runs independently. When all subprocesses are finished, the whole process is finished. The following procedures directly support this kind of parallel processing:

• SDO_GEOR.generatePyramid
• SDO_GEOR_AGGR.mosaicSubset
• SDO_GEOR_UTL.emptyBlocks
• SDO_GEOR_UTL.fillEmptyBlocks
• All the procedures in SDO_GEOR_RA Package Reference
• All the procedures in SDO_GEOR_IP Package Reference

Through the SDO_GEOR_AGGR.mosaicSubset procedure, other types of parallel operations are supported. These include parallel compression and decompression, parallel copying or change format copying, parallel subsetting, parallel reprojection, and parallel rectification. See Parallel Compression, Copying, and Subsetting for more information.

Imagery and raster data are typically very large, so the preceding operations can be time consuming. Therefore, when using multi-CPU or multicore servers, always consider using parallel processing to improve the performance.