Case Study: Hulu achieves 800% query performance improvement and scales to serve 4 billion videos with Pivotal's Redis solution

Hulu Leading video company scales to serve 4 billion videos with 800% query performance improvement

Hulu, a leading U.S. streaming service with millions of viewers and subscribers, faced a scaling crisis in 2012 when its viewer-history system outgrew a MySQL + Memcached design. MySQL could not handle the write volume, Memcached couldn’t be replicated so reads were concentrated on a single shard in one datacenter, and the result was high latency, degraded query performance and no high-availability strategy.

Hulu replaced the bottleneck with a Redis-based architecture (sharded by user_id and replicated across datacenters), built a custom failover mechanism, and used Cassandra for durable writes while caching active user data in Redis. The new design provided linear scalability and high availability, cut read latency dramatically (e.g., east-coast reads from ~120 ms to <15 ms for most requests), enabled ~10k QPS capacity, and delivered an estimated 800% improvement in query performance.

Hulu