Case Study: Ola Cabs achieves low-latency, high-throughput production and reduced infrastructure costs with ScyllaDB

Two Years of Using ScyllaDB in Production

Ola Cabs faced the classic real-time challenge of ride-hailing: large, unpredictable spikes in demand that require a database capable of very high throughput, low latency, and multi‑AZ/high‑availability operation with minimal maintenance. Seeking an open‑source, non‑JVM alternative to Cassandra, Ola evaluated and began adopting ScyllaDB (Scylla) starting in 2016, initially in passive write mode and then for both reads and writes as the product matured.

ScyllaDB delivered a native, low‑latency, NVMe‑optimized datastore plus tuned AMIs and Prometheus metrics that sped debugging and sizing decisions. Ola moved key workloads onto ScyllaDB, achieved target latencies (sub‑5 ms 99+ percentile for many use cases), avoided a separate caching tier (reducing hardware costs), and ran stable clusters in production for two years—while also benefiting from reduced JVM overhead, better Go‑driver performance, and faster performance tuning through shared metrics.

Ola Cabs

Soumya Simanta

Architect