how to get the nearest 100 points from one million data records quickly?

Question

given one point A, get the nearest 100 points from one million data records

• database is MySql
• one million records of latitude and lon

Answer 1

http://en.wikipedia.org/wiki/Geohash might be a quick way to speed up the average case, but worst case behaviour would still be bad. The article suggests that you index by geohash and, on a query, retrieve all points in a bounding box that amounts to a geohash prefix. If the bounding box is small and you find a match in it closer than any point outside the bounding box then you have succeeded quickly, but neither of these things might be true.