Taming P99s in OpenFGA: How We Built a Self-Tuning Strategy Planner
Summary
Auth0's OpenFGA post describes a self-tuning strategy planner that uses Thompson Sampling to minimize P99 latency in graph traversals by adapting per-subgraph strategies. It explains Bayesian priors via Normal-Gamma and continuous runtime learning, with production results showing significant P99 reductions and occasional retention of legacy paths for certain distributions.