At ScyllaDB, P99 is part of our DNA. Predictable low latency has been a hard requirement ever since we launched in 2015.
In this keynote, ScyllaDB co-founder Dor Laor shares the lessons learned from various challenges along our P99 quest. First, we take a technical look at specific ScyllaDB optimizations (hitting our initial latency targets, dealing with 200+ node clusters, network topologies, goodput/badput, connection storms, etc.). The focus then shifts to P99 concerns shared across the community of latency-focused products (e.g., believable benchmarking).
