Case Study: LaunchDarkly cuts feature flag initialization time with Fastly Compute

LaunchDarkly uses Compute@Edge to ensure speed and security anywhere in the world

LaunchDarkly, a pioneer in feature management and feature flags, needed to improve the speed and reliability of flag evaluation for customers using its SDK in polling mode, especially as it evaluated 20 trillion feature flags per day. While it had already used Fastly’s edge cloud platform for purging and shielding, some regions still saw initialization times above one second, creating latency concerns for global users.

LaunchDarkly implemented Fastly Compute to move flag evaluation closer to users at the edge, creating Flagbearer@Edge. With Fastly, the company cut average initialization time to 25 milliseconds, achieved a 98% cache hit rate, and now serves up to 100% of production polling traffic on Compute, while also benefiting from Fastly Enterprise Support and stable, low-maintenance infrastructure.

LaunchDarkly

Andrew Brown

Senior Software Engineer