Snowflake Cost Optimization Best Practices

4. Use Warehouse Configuration properly: For end-user warehouses where performance is a priority, set the MAX_CLUSTER_COUNT = 5 and SCALING_POLICY = ‘STANDARD.’ This will automatically allocate additional clusters as the concurrent workload increases. However, put the MAX_CLUSTER_COUNT to the smallest number possible while controlling the time spent queuing. With a SCALING POLICY of STANDARD, avoid setting the MAX_CLUSTER_COUNT = 10 (or higher) unless maximizing performance is a much higher priority than controlling cost.
5. Using Auto_Suspend: Only set the AUTO_SUSPEND to zero if you are confident you are aware of the potential risks of continuous spending and take steps to suspend warehouses when not needed.
6. Update the Query Timeout Default Value: Snowflake has a default query timeout value of 48 hours (172800 seconds), meaning that queries running longer than 48 hours will be automatically terminated. Updating the query timeout value can help reduce Snowflake costs by preventing long-running queries from consuming excessive compute resources.

Set a STATEMENT TIMEOUT IN SECONDS for every warehouse based on the warehouse size to limit the risk of overspending on an individual query. 

7. Using Resource Monitor Tool:  It is important to set up resource monitors to keep the Snowflake usage within budget. These monitors can be set up to send alerts and/or stop WH automatically when consumption levels reach predefined thresholds.  

Setting the notification thresholds below your hard limit is recommended to avoid running out of capacity. These alerts can be selected at an Account level (comprising all WHs) or for individual warehouses.