Neon's autoscaling feature for Postgres databases has graduated from beta to general availability, including for Free plan users (up to 2 CPU, 8 GB RAM). The system dynamically adjusts CPU and memory based on workload, eliminating the need for manual instance resizing and reducing overprovisioning costs. Architecturally, it uses NeonVM for VM management within Kubernetes, autoscaler-agents that poll metrics every 5 seconds, a modified Kubernetes scheduler for pre-approved upscaling, and a local file cache acting as a resizable extension of Postgres shared buffers. Users set min/max autoscaling limits via the console or CLI, with scale-to-zero support for non-production branches.
2 Impressions