Deployment set routing #

Starburst Galaxy can handle spikes in workload demand by autoscaling a cluster, but a single cluster still has finite processing capacity. If a cluster becomes overwhelmed with concurrent workloads, you might experience performance degradation and slower response times, even with autoscaling enabled.

To handle this type of heavy demand on a cluster, you can use deployment set routing. Deployment set routing launches multiple Trino deployments as replicated clusters. Replicas are identical to the original cluster, use the same cluster URL, and do not need to be manually managed.

When you have one or more replicated clusters, Galaxy routes queries across multiple coordinators using a routing mechanism called smart load balancing. Smart load balancing continuously evaluates workload distribution and routes queries in real time based on system load. This reduces query latency, minimizes queueing, and utilizes resources more efficiently than traditional load balancing methods. Smart load balancing is automatic and does not require any configuration.

You can use deployment set routing alone, or in combination with user role-based routing. No changes to routing rules are required. You can create up to eight replicated clusters, either when you create a new cluster or edit an existing cluster.

Limitations #

The limitations of deployment set routing are:

• Auto-suspending is not supported. By extension, cluster schedules are not supported.
• Autoscaling is not supported.
• The Trino UI is not supported.