[mapserver-users] Help (tips&tricks) on raster dataset optimization

[BrainDrain]

Hi, friends. 
I need an advice to choose some way to optimize my raster dataset.
Source data: 39 geotiff files - satellite images of a city, each one is
aprox. 600 mb uncompressed rgb/~250 mb lzw. To speedup rendering in
different scales I can:
1) - transform source files into internally tiled files with gdal_translate
(lzw or deflate? lzw is faster?)
- create virtual mosaic (.vrt) with gdalbuildvrt 
- create internal overviews with gdaladdo on vrt (lzw or deflate for
overviews compression? can this method differ from method we choose earlier)
- connect to result dataset with MS (no minscale/maxscale manipulations
required? can I connect to directly to vrt or I need to create tileindex
with gdaltindex (internal overviews supported in this case by MS?)?)
2) - create external overviews (pyramid levels) of source files
- break this overviews and source files into tiles
- create tileindexes for each level of details (pyramid level) with
gdaltindex
- connect to result tileindexes with MS using different minscale/maxscale
conditions
3) - convert source files into ecw's (what compression level in % is
reasonable? good looking (vs loosless))
- connect to result dataset (can I use tileindex for ecw's?)
Have someone experience in comparing those ways?
Thanks!
-- 
View this message in context: http://n2.nabble.com/Help-tips-tricks-on-raster-dataset-optimization-tp3825869p3825869.html
Sent from the Mapserver - User mailing list archive at Nabble.com.