Serving Driver-partners Data at Scale Using Mirror Cache

This post introduces 'mirror cache', an in-memory local caching solution for the Drivers Data service. It addresses low cache hit ratios by replicating data across all live instances of the service. The design goals include supporting LRU cache, active invalidation, best-effort replication, and auto-scaling. The implementation uses Ristretto for the in-memory cache and a replicator component with a Membership Store, Notifier, and gRPC Server for data mirroring. The new architecture resulted in a local cache hit ratio of 0.75 and a 5% drop in database calls.