A compute cluster is similar to a multinode database instance.
Compute clusters belong to compute groups. You can have different compute groups for different types of work (see Standard Compute Clusters, Analytic Compute Clusters, and Use Cases for Compute Clusters).
Compute clusters do not have their own permanent storage. The primary cluster and multiple compute clusters access data in Object Storage. The primary cluster distributes the workload to a single compute cluster, selected based on user membership in a compute group.
Because compute clusters are independent and get work from primary clusters, they can be added or removed as needed. Failure of one compute cluster does not disrupt work on other compute clusters.
The number of compute clusters in an Database Engine 20 depends on the number of databases sharing the account and their resource needs. See the description of compute_map in CREATE COMPUTE PROFILE Syntax Elements.