Fastly Case Studies Split's Rapid Scaling with Fastly's Edge Cloud Network
Fastly Logo

Split's Rapid Scaling with Fastly's Edge Cloud Network

Fastly
Analytics & Modeling - Computer Vision Software
Infrastructure as a Service (IaaS) - Cloud Middleware & Microservices
Equipment & Machinery
Maintenance
Product Research & Development
Traffic Monitoring
Transportation Simulation
Cloud Planning, Design & Implementation Services
System Integration
Split, a global provider of a feature delivery and experimentation platform, was seeking a scalable and reliable edge cloud network to deliver feature flags to its customers. The company's customers use these feature flags to introduce and roll back changes swiftly, thereby reducing the median time to recovery. However, Split was facing challenges with conventional Content Delivery Networks (CDNs), experiencing purge times ranging from 10 to 30 minutes, which was unacceptable for their business model. Furthermore, as a company with a worldwide customer base, Split needed to maintain high uptimes in every location, a task that was proving difficult due to the potential for human error and the challenges of performance at scale.
Read More
Split is a global company that provides a robust feature delivery and experimentation platform. The platform serves billions of feature flags each day, monitoring and measuring the impact of newly released features so customers can link changes to key business objectives. Since its founding in 2015, Split has grown exponentially, from delivering 1 billion feature flags per month in 2017 to 1 trillion in 2020. The company's customers use feature flags to introduce and roll back changes swiftly, reducing the median time to recovery. Split has a worldwide customer base and is committed to maintaining high uptimes in every location.
Read More
Split chose Fastly's edge cloud network for its scalability and reliability. With Fastly, Split achieved sub-second purge times in most cases, a significant improvement from the 10 to 30 minutes purge times they were experiencing with conventional CDNs. Fastly's edge cloud network also allowed Split to maintain high performance even during traffic peaks by delivering flags from the edge and strategically shielding their origin. Fastly's global network enabled Split to maintain high uptimes in every location, with Service Level Agreements (SLAs) offering uptimes as low as 99.99%. Fastly's developer tools and high-touch support also helped Split grow, providing guidance and support as the company developed new features or modified code. Split also utilized Fastly's real-time observability integrations with Sumo Logic and syslogs for real-time insights into their traffic.
Read More
Fastly's edge cloud network has enabled Split to deliver feature flags to its customers more efficiently and reliably. The sub-second purge times have significantly improved the company's service delivery, ensuring that customers can introduce and roll back changes swiftly. The high performance during traffic peaks has ensured that Split can maintain reliably high performance even as traffic peaks. Fastly's global network has also allowed Split to maintain high uptimes in every location, ensuring reliable service delivery to its worldwide customer base. Furthermore, Fastly's developer tools and high-touch support have been instrumental in Split's growth, providing the necessary guidance and support as the company developed new features or modified code. Fastly's real-time observability integrations have also provided Split with valuable insights into their traffic, enabling them to manage unexpected spikes and maintain smooth traffic flow.
Split achieved sub-second purge times with Fastly, a significant improvement from the 10 to 30 minutes purge times with conventional CDNs.
Fastly's edge cloud network allowed Split to maintain high performance even during traffic peaks, with only 54 out of 120,000 requests per second not going to Fastly.
Fastly's global network enabled Split to maintain high uptimes in every location, with SLAs offering uptimes as low as 99.99%.
Download PDF Version
test test