Ideas how to improve import speed when using a a lot of large GeoTiffs for elevation info

I’d love to get ideas from you people, on how to improve import speed when using a local geotiff for elevation information.

We have world-wide elevation information separated out into a looooot of .tif files. Those are all combined using one .vrt file and I wrote an importer that uses gdal-Java bindings to read elevation information.
This works fine on my local machine where I only have a small geotiff for testing. But on our server, we have those huge geotiff described above.

Now reading 100 elevation infos locally takes about 1ms (or less… we didn’t do suuuuper accurate measuring) but on the server this takes about 90ms. So massively slower.
I suspect the .vrt file (with it’s more that 1000 sub-geotiffs) to be the reason. When going directly into the linked geotiff on the server it’s also a lot faster…

Did anybody here already solve that problem? Or do you have ideas on how to solve it?

What I was thinking of trying is to

Why not using the tif files directly without the vrt container?

What I was thinking of trying is to

There seems to be missing something :slight_smile:

btw: to avoid excluding female developers I suggest using e.g. “people” instead of “guys” :slight_smile: (I also had to learn this as I’m not a native speaker)

Hey @karussell thank you very much for your quick help!
I also meant the female developers of course :slight_smile: So all the people :slight_smile: sorry

Ok… so the problem is that we have 1000+ geotiffs and they are layered. And all with arbitrary names, so I can’t convert a coordinate to a filename. When we get a better tif for a region, we use that one. So f.e. we’d have one not so high resolution one for all of Europe, but then a higher resolution for f.e. Germany. When using the vrt, it’s smart enough to know that the Germany layer is layered above so it needs to look at that geotiff and therefore gives me the higher resolution result.

Also the gdal-jni does not provide all the power gdal has to offer. Cause gdal also knows how to look that up… So I’d have to do that calculation manually I guess? And as I don’t know thaaat much about GIS that’d be kind of hard for me…

I was hoping somebody already encountered this problem?!?