I have a large dataset of aerial images consisting of 8000 TIFFs where single images are no larger than 0.5G and all have have .ovr pyramid files with them.

Previously I have served this dataset as imageMosaic, but I'm wondering should I use imagePyramid instead. At what point is imagePyramid better than mosaic of TIFF+ovr or is it ever?

As far as I understand GeoServer imageMosaic supports ovr-files so is there any benefit of creating pyramid folder structure?


1 Answer 1


Most (all?) operating systems will baulk at opening 8000 files at a time, so if your user is zoomed out and tries to view your entire set of imagery at the same time it will fail. An image pyramid avoids this by providing a smaller set of lower resolution images at the upper zoom levels.

0.5 Gb files sound quite small (remember there is an overhead to opening a file) and I would look to combining the base images into bigger, tiled, compressed images with overviews and then create a pyramid of them.

As always you need to experiment to find the "optimum" balance and I would follow the tips in GeoSolutions' video and Paul Ramsey's compression for dummies blog post to see what a difference preparing your data can have.

  • Thank you for the links, I'll look them up. My original post didn't mention that the images are YCBCR jpeg tiffs so as per Paul Ramsey's post they should be optimal. Though I haven't checked if they have internal tiling and/or overviews. By pyramid do you mean external pyramids (the .ovr files) or pyramid used by GeoServer (which to my understanding is a folder structure)?
    – jfp
    Nov 1, 2019 at 9:58
  • I mean GeoServer's pyramid extension
    – Ian Turton
    Nov 1, 2019 at 10:15
  • What about overviews: are they still useful after the pyramid is created (using gdal_retile.py). Isnät overviews and pyramid basically the same thing?
    – jfp
    Nov 1, 2019 at 13:31
  • overviews can still be used at the individual levels of the pyramid, but are less useful in a well designed pyramid
    – Ian Turton
    Nov 1, 2019 at 15:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.