Breaking up a monolith: How we’re unwinding a shared database at scale | Datadog

What this post added

This post details Datadog's strategy and execution for unwinding a large, shared monolithic database into independently owned instances. It outlines a three-phase plan: logical separation (identifying data domains and responsibilities using DBM query samples and engineer collaboration), access reduction (minimizing direct database access by enforcing API/gRPC service usage and implementing a homegrown PG Proxy for flexible routing), and physical separation (migrating data to new, independently owned instances managed by the OrgStore platform). The post emphasizes the importance of platform investments like Rapid (for service building) and OrgStore (for managed databases) in making such migrations feasible and safe.