PointFive
Back to Knowledge Base
Compute

Underutilized Snowflake Warehouse

Simar Arora
January 15, 2025·4 min read

Explanation

Underutilization occurs when workloads run on oversized warehouse instances. For instance, a task needing a Medium warehouse may execute on a Large or Extra Large one. This causes excess credit consumption without proportional performance gains. Root causes include initial over-provisioning without later reassessment or pursuit of marginal speed improvements that don't justify higher costs.

Relevant Billing Model

Snowflake's pricing structure bases charges on warehouse size and active runtime duration. Larger warehouses consume credits at higher rates per second than smaller configurations. Organizations should select appropriately-sized warehouses aligned with actual workload demands.

Detection

  • Examine historical query activity per warehouse to evaluate concurrency patterns and peak usage
  • Assess whether queries experience queuing or if the warehouse remains frequently underused
  • Compare credit consumption against query volume and complexity metrics over a representative period
  • Validate with engineering teams whether downsizing is feasible without breaching SLAs

Remediation

  • Downsize the warehouse to a smaller tier supporting adequate performance and concurrency
  • Establish recurring reviews assessing sizing based on usage patterns and workload changes
  • Coordinate with teams to confirm SLA compatibility before implementing changes
Back to Knowledge Base