Case Study: Natura achieves dramatic performance gains and 50%+ infrastructure cost savings with ScyllaDB

Natura Achieves Beautiful Results with Scylla

Natura, the largest Brazilian cosmetics company, ran a Kafka/Spark streaming architecture and relied on Cassandra for key‑value workloads but faced significant performance and cost challenges—high latencies largely due to the JVM/Garbage Collection, large hardware requirements (11 primary disks on a 5‑node ring), and operational overhead—so they evaluated alternatives including ScyllaDB.

Natura migrated to ScyllaDB (Scylla) and saw dramatic improvements: overall batch processing dropped ~10% with some jobs going from 6 hours to under 10 minutes, average processing time fell from 76 ms to 6 ms, and 95th‑percentile latency improved from 220 ms (Cassandra) to ~500 µs (Scylla). ScyllaDB also reduced required primary disks from 11 to 2, cutting hardware costs by over 50%, and provided strong developer support and the ability to scale by cores rather than nodes.

Natura

Felipe Moz

Big Data Engineer