MegatronX: The powerful geospatial data production system

メガトロンX

Who or what is Megatron X? Nobody knows according to the wiki. All that is known is that his battle capabilities are off the charts.

Inspired by the original Megatron X, one of the ERDAS IMAGINE development team  (ThePixel) decided to name their super powerful computer on loan from HP as MegatronX. The two have a lot in common, this computer's configuration is of the charts too.

• Model         :  HP Z820 Workstation
• RAM           : 128 GB total (16x 8 GB modules) DDR3 Quad, 
• RAID          : 15 TB RAID 1 (7200 RPM SATA HDD)
• SATA SSD   : 256 GB
• ZTurbo SSD : 750 GB 
• Processor    : 8 cores @ 2.9 Ghz (16 cores) - 2x Intel Xeon E5 2690 

MosaicPro is one of the prominent data production tool in ERDAS IMAGINE. It is used to mosaic/stitch remote sensing images. The data desired to be stitched into one single file could be of the order of 500+ GB of compressed data. With the 2014 release of IMAGINE, several applications were coverted into a 64 bit application removing the limitation of 32 bit application which would limit the use of RAM to 2GB. Apart from it, Mosaic Pro is constantly improved to provide a faster and high quality output. These experiments are generally memory and storage intensive. MegatronX has been purposed for benchmarking and experimenting in this regards. On this blog post will publish some of our findings as they become available.

Mosaicing the entire state of Oregon

Input Data    : 1909 individual images in JPEG compressed TIFF with total file size of 700+ GB.

Output Data  : Single Mosaiced image in ECW compressed with total file size of 168+ GB.

Input drive       : SATA

Output drive     : SATA

Time to mosaic : 271 h 57 m 06 s

Note

• In past before the 64 bit conversion mosaicing this dataset was impractical
• On a regular machine this dataset would take several weeks

Mosaicing the entire county of Cherokee 

Input Data    : 2094 individual images in IMG format with total file size of 162+ GB.

Output Data  : Single Mosaiced image in IMG format with total file size of 182+ GB.
Output Data (alternatively) : Single Mosaiced image in ECW compressed with total file size of 7+ GB.

Input drive       : SATA
Output drive     : SATA
Time to mosaic : 271 h 57 m 06 s

Note

• In past before the 64 bit conversion mosaicing this dataset was impractical
• On a regular machine this dataset would take several weeks

Least Cost Path using Spatial Modeler

The Whats and Whys?

Compute the least cost path between two points/location in Rabun county. 

The first point/location we picked was Rabun County Jail and the second one was somewhere in Rabun Bald. We picked these locations to re-enact a hypothetical prison break/escape scenario, for some fun, but eventually got dropped and we did not to continue too deep with the theme. The example though stayed with us and is used extensively by my team in our automated testing suites for ERDAS IMAGINE one of the flagship products of Hexagon Geospatial. 

This blog post is my attempt to use the example as a tutorials to show how to use the Least Cost Path operator in Spatial Modeler.

Using the Operator

Create a Cost Map

Video still to come...

Filter and Rasterize the to/from points from a larger data

Video still to come...

Whole big model

Video still to come...