Our Approach to High Availability

High availability (HA) in managed databases refers to the system's ability to remain operational and accessible "for a high percentage of time," even during failures. Neon, a managed database service, ensures HA by separating storage and compute layers. Compute failures are resolved quickly as stateless processing nodes can be replaced immediately. Storage is designed with redundancy and ultra-high durability through the use of cloud object storage, automatically replicating and distributing data across multiple nodes for failure tolerance. Neon's Safekeepers handle Write-Ahead Log (WAL) replication with redundancy across multiple AZs, while Pageservers act as a fast-access disk cache for compute. The primary copy of all data resides in cloud object storage, providing ultra-high durability.