Amazon Redshift Costs: Pause, Resume and Idle Clusters

Node hours, storage, and the pause

A provisioned Redshift cluster is billed an hourly rate per node for as long as it runs. RA3 managed storage is billed separately at a GB-month rate. Redshift Serverless is priced in RPU-hours per second with no charge while idle.

AWS lets you pause and resume a provisioned cluster, and while it is paused on-demand billing is suspended and only storage is charged. The console can create a recurring pause and resume schedule. A few things to know before you rely on it:

• Pausing takes a snapshot and terminates running queries, and connections are unavailable while paused.
• EC2-Classic clusters, HSM clusters and clusters with automated snapshots turned off cannot be paused.
• Node IP addresses can change on resume, and query performance can dip while the cluster re-hydrates.
• A Redshift admin secret in AWS Secrets Manager keeps billing during the pause.

What ZopNight does for Redshift

ZopNight discovers provisioned clusters with the rest of your AWS estate and shows their cost as rack rate or billed cost. The Idle Redshift Cluster rule fires when 30 days of database connections sit below a small floor, with CPU used as a guard so a cluster doing real work with few sessions is left alone. The recommended action is a manual, reversible pause, and the saving is priced on the compute share of the cluster’s cost, leaving the storage residual in place.

ZopNight can also put provisioned clusters on a schedule, with the same schedules, groups and overrides as VMs: the stop pauses the cluster and the start resumes it. The Idle Redshift Cluster recommendation itself is advisory: ZopNight shows the evidence and the saving, and you apply the pause or a schedule yourself. The remediation steps also point to Redshift Serverless for infrequent workloads.

Related

See the Redshift fact sheet and the Redshift pause and resume notes. For the idle method across services, read idle detection.