[Benchmarking] Testing mosaics, how fine grained?

[Jeff McKenna]

Andrea Aime wrote:
> Hi,
> mosaic wise I was wondering if we should try out more fine grained
> setups? At the moment we have these eight 1.8GB tiles set, I was 
> wondering how realistic it is, and what indications we
> should provide to people trying to setup raster data to the
> respective servers.
> 
> E.g., will a mosaic made of 128 files perform worse (I guess so).
> By how much? What is the gain in preprocessing the files so they
> get close to the bigtiff limit? Any improvement in going fully
> bigtiff (ok, this is something I won't be able to answer with
> GeoServer, but the progression from 128 to just 8 would be interesting.
> 
> I guess it also matters a lot how the mosaic is used, free zoom
> access or clamped to nearly the native resolution?
> 

Since we'll test tile indexes on the vector side (roughly 250 counties), 
I think it would be great to also test tile indexes on the raster side 
with a large number in the hundreds...it's the only way we can do a real 
world test.

-jeff