This page describes operational scaling for SurrealDB Cloud: changing instance size on Start, adding compute nodes on Scale, and using observability data to time changes safely.
Resizing Start instances
On the Start plan, resize when monitoring shows a steady mismatch between provisioned resources and workload — for example, sustained high CPU or memory after optimising queries. In Surrealist or the Cloud dashboard, open the instance, choose a new instance type or storage size, and apply the change. A short period of elevated latency or reconnects may take place while the platform adjusts capacity, so plan changes outside peak traffic when you can.
Scaling Scale clusters
On the Scale plan, clusters run on SurrealDS with a minimum of three compute units. Increase compute units to add query nodes and adjust storage as the dataset grows. Horizontal scale improves throughput and fault tolerance, allowing the cluster to keep serving when one node is unavailable.
Where your plan supports read replicas or related features as they roll out, adding read capacity can further improve availability. Confirm replication lag stays within application tolerances after scaling out.
Monitoring before and after
Use metrics (CPU, memory, storage, connections, query latency) and logs to confirm whether scaling solved the bottleneck or merely moved it (for example, from CPU to disk). If utilisation remains low after a large resize, consider stepping down to avoid overspending.
Together with backup and network settings, scaling is part of routine operations—see Monitoring overview for related dashboards and signals.