This paper addresses the non-linear isomorphic Dvoretzky theorem and the design of good approximate distance oracles for large distortion. We introduce and construct optimal Ramsey partitions, and use them to show that for every ε ∈ (0,1), any n-point metric space has a subset of size n 1-ε which embeds into Hilbert space with distortion 0(1/ε). This result is best possible and improves part of the metric Ramsey theorem of Bartal, Linial, Mendel and Naor , in addition to considerably simplifying its proof. We use our new Ramsey partitions to design approximate distance oracles with a universal constant query time, closing a gap left open by Thorup and Zwick in . Namely, we show that for any n point metric space X, and k > 1, there exists an O (k)-approximate distance oracle whose storage requirement is O(n1+1/k), and whose query time is a universal constant. We also discuss applications to various other geometric data structures, and the relation to well separated pair decompositions.