How we made feature flags even faster and more reliable

PostHog undertook a significant overhaul of its feature flag evaluation service by rewriting it from Django to Rust to address scalability challenges and improve performance and reliability. The previous Django-based system faced bottlenecks due to technical debt, inefficient database queries, and reliance on PgBouncer for connection pooling, which led to high compute costs and fragile reliability. By transitioning to Rust with the axum framework, PostHog achieved a 21x increase in throughput and reduced latency significantly, with p99 latency dropping from 904ms to 85.4ms. The rewrite involved moving flag evaluation logic from SQL to application code, implementing app-level cohort caching, and eliminating PgBouncer, leading to a 68% reduction in compute costs while maintaining the same traffic capacity. This architectural simplification improved reliability, eliminating feature flag outages and ensuring that the service remained responsive even under database pressure, thus reinforcing the principle that the best optimization sometimes involves doing less.

No tracked trend matches for this post yet.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.