Case Study: Split achieves 1000x scale and sub-second feature delivery with Fastly

Split scales 1000x with Fastly to deliver trillions of feature flags and experiments each month

Split needed a scalable, reliable edge cloud network to deliver feature flags and experiments to its global customers, but traditional CDNs were too slow, with purge times of 10 to 30 minutes. As Split grew rapidly and began serving billions of flags per day, it turned to Fastly to support instant rollbacks, high uptime, and performance at scale.

Fastly implemented edge delivery with Instant Purge, Origin Shielding, global load balancing, real-time logging, Terraform integrations, and Image Optimizer. The result was sub-second purge times, protection during traffic spikes, and reliable delivery even at 120,000 requests per second, with almost all traffic served at the edge. Fastly also helped Split improve SEO, reduce human error, and scale from 1 billion feature flags per month in 2017 to 1 trillion in 2020.

Split

Patricio “Pato” Echagüe

Chief Technology Officer and Co-founder