We use the S2 Spherical Geometry Library for mapping points and regions into single dimension 64bit CellID representation, indexable as an integer index.
Q2: Any ballpark on the number of queries that could be handled by this (as compared to say regular read requests)?
We are trying to maintain the right balance of query execution while maintaining a high write rate for Geo data. We know of optimizations that could further accelerate query performance, but before we go down that path we wanted to get a better idea of real world performance and needs. This is where the community comes in. We know your mileage will vary (number and size of expected return value sets etc.) - we want to understand what that looks like in the real world.
