Case Study: Outbrain achieves 4x latency improvement and 33% smaller cluster footprint with ScyllaDB

Outbrain manages 2 billion records with persistence and always-on availability with Scylla

Outbrain, the world's largest content discovery platform with 557 million unique visitors and 250 billion personalized recommendations per month, faced performance and cost challenges with its Apache Cassandra–backed document database that stores 2 billion records. Cassandra’s latency forced Outbrain to use an in-memory Memcache layer, adding complexity and consistency headaches, while throughput demands required very large, costly clusters. To find a better option, Outbrain evaluated ScyllaDB by running a Scylla cluster side‑by‑side with their existing Cassandra deployment.

Outbrain deployed a ScyllaDB cluster in the same data center, serving the same traffic without an in‑memory cache and with clients reading and writing to both databases while monitoring latency and throughput. ScyllaDB handled the load previously served by the Memcache + Apache Cassandra combination and delivered a 4x latency improvement and a one‑third reduction in cluster size, simplifying operations and lowering costs.

Outbrain