On 3/20/2018 1:40 PM, sav123 wrote:
Kompza covered the variables you mentioned, but I if you're looking at
benchmarking, I recommend you set up the full rendering stack rather
than trying to generate queries yourself. The latter can be tricky to
get right, and you have to handle parallelism the same as the full stack
to get meaningful results. It's just easier to use renderd and feed it a
list of tiles with render_list than to write code that will take that
list, generate SQL, and run it.

This doesn't answer the question.
If you read the carto file, you will see that there are natural criteria , not only geometry. Indeed, it is a bad idea to partition on coordinates, unless if a request applies to a small region.

I'm not speaking of partitioning, but of how to best benchmark performance, which is what you're planning to do to see if partitioning is worthwhile.
_______________________________________________
dev mailing list
dev@openstreetmap.org
https://lists.openstreetmap.org/listinfo/dev

Reply via email to