How Agoda Scaled Its Price Cache with DragonflyDB

This post details the migration of Agoda's Price Cache from a sharded MSSQL database to DragonflyDB. It outlines the challenges faced with scaling MSSQL, the selection criteria for DragonflyDB (multi-threading, Redis compatibility, built-in expiry, internal expertise), and performance validation through load testing. The migration strategy is described in four phases: planning/preparation (including initial hot data storage and later full dataset sharded deployment), parity checking (comparing supplier count and price-data length), gradual rollout via A/B experimentation, and post-migration optimizations like dual-cluster high availability and automated failover. The post quantifies performance improvements in write throughput and latency, and read latency under production load.