Hello everyone,
I'm working on analyzing cost projections for Serverless SQL Warehouses using system tables, and Iād like to share a visualization approach Iām using to highlight some key differences between classic and serverless SQL warehouses. (Look SS)
Graph 1: Query Execution Timeline
X-axis: Time of day
Y-axis: Date (covering the last month)
Details: This graph shows when queries were triggered across different warehouse_ids and how long each query took to execute.
Graph 2: Classic SQL Warehouse Cluster Duration
X-axis: Time of day
Y-axis: Date (covering the last month)
Details: This graph illustrates the cluster duration times for queries executed on classic SQL warehouses for selected warehouse_ids.
Graph 3: Serverless SQL Warehouse Cluster Duration
X-axis: Time of day
Y-axis: Date (covering the last month)
Details: This graph shows the cluster duration times for queries executed on serverless SQL warehouses for selected warehouse_ids.
Key Insight:
The goal of this comparison is to demonstrate that classic SQL warehouses remain active even when no queries are running, leading to potentially higher costs. In contrast, serverless SQL warehouses spin up immediately when a query is triggered and shut down right after execution, offering a more cost-efficient model.
I look forward to hearing your thoughts, suggestions, and queries on how to obtain this visualization and how to improve this analysis and visualization.
Thanks in advance.
Hi @mstfkmlbsbdk, great analysis.
However, as you mentioned, the classic warehouse is not terminating even though there are no active queries on it. The reason for that is while creating a classic warehouse, by default it takes 45 minutes to terminate automatically if there is no active queries. You can set it to minimum 10 minutes which is the default for serverless, and can save the cost.
