Modern distributed systems hide significant complexity behind simple user interactions. In this talk, I will give a high-level look behind several of our services running on Kubernetes in Azure, including SDS, Entity, and Events stores. We’ll explore how SDS scales by partitioning workload across Azure Storage, and how RocksDB is extended with cloud-backed durability to support high-throughput graph workloads. Rather than deep implementation details, the focus will be on the architectural patterns and trade-offs that make these systems scalable, resilient, and efficient.