Efficiently handle rasters that exceed heap and main memory size.
Current implementation ueses default DataBuffer backend for WriteableRaster as backend for GridCoverage. DataBuffer uses a plan primitive[] array for data storage. That is, a raster can have max. 2G = 46K x 46K raster elements and allocates 2GB x 8 = 16GB size on heap.
Goals
process raster more independently of actual memory and heap size
allow to process rasters > 2GB on 32bit systems
maybe use memory-mapped files as persistent storage, avoid write/encode after processing
TODO
[ ] eviction: limit mapped files on 32bit systems
[ ] memory sensitive on 32bit systems (soft references and check for allocation error)
[ ] check whole-file version for 64bit systems
[ ] stride/tile version
[ ] check memory-mapped files as persistent storage
Efficiently handle rasters that exceed heap and main memory size.
Current implementation ueses default DataBuffer backend for WriteableRaster as backend for GridCoverage. DataBuffer uses a plan primitive[] array for data storage. That is, a raster can have max. 2G = 46K x 46K raster elements and allocates 2GB x 8 = 16GB size on heap.
Goals
TODO