Understanding what levels to use in gdaladdo

Question

I am trying to understand how tiles overviews and levels interact using GDAL so i know how to use it appropriate and optimize my rasters to use either in QGIS and/or geoserver.

So just to verify my understanding. TILED=YES (using ex gdal_translate). Tiles the internal file structure of instead of reading it linear. OVERVIEWS (using gdaladdo). Creates a pyramids. Can be created on individual GeoTiff or on a vrt. My main issue is to understand levels.

For example

gdaladdo -ro -r 
--config COMPRESS_OVERVIEW JPEG 
--config PHOTOMETRIC_OVERVIEW YCBCR 
--config INTERLEAVE_OVERVIEW PIXEL 
––config GDAL_TIFF_OVR_BLOCKSIZE 256 
--config JPEG_QUALITY_OVERVIEW 75 
mosaic.vrt 2 4 8 16 32 

Say I have a mosaic with aerial images for a large region, say a country.
What levels should I use to be able to render the entire country fast (as a guideline)? What does 32 mean and how do I know where to stop? 64, 128? and how does it relate to the blocksize to use?

But if I understand correctly using these pyramids in geoserver would not work as it cannot read vrt files. Then I need to tile them with gdal_retile?

Answer 1

It depends how big your last level of overviews is.

Is it a 100000x100000 pixel image? Then it would probably be too big to be useful (a client or server using the file would need to read quite a few bytes to generate its image or analysis). Is it 1000x1000? Then it would be ready to use for zoomed-out views.

The resolution corresponding with the level depends on your input. If your input is 512000x512000, then level 8 would be 64000x64000. If it was 10000x10000, then level 8 would be 1250x1250.

Modern GDAL will generate all reasonable levels if you just leave out the level numbers in your commandline.