topic Is it more performant to run optimize table commands on a serverless SQL warehouse or elsewhere? in Data Engineering

Is it more performant to run optimize table commands on a serverless SQL warehouse or elsewhere?

ac0 — Fri, 10 May 2024 03:02:51 GMT

Is it more performant to run optimize table commands on a serverless SQL warehouse or in a job or all-purpose compute cluster? I would presume a serverless warehouse would be faster, but I don't know how to test this.

Re: Is it more performant to run optimize table commands on a serverless SQL warehouse or elsewhere?

-werners- — Thu, 16 May 2024 14:07:35 GMT

I am not so sure a serverless cluster would be faster than a photon-enabled interactive/job cluster (assuming the same scale).
You could do a test with a deep clone of the table and apply optimize on both using different clusters.

Re: Is it more performant to run optimize table commands on a serverless SQL warehouse or elsewhere?

Yeshwanth — Thu, 16 May 2024 15:35:43 GMT

@ac0 Good day!

Serverless SQL warehouses are likely to execute "optimize table" commands faster than job or all-purpose compute clusters due to their rapid startup time, quick upscaling for low latency, and efficient handling of varying query demand. They also downscale quickly to minimize costs when demand is low. To compare execution speeds, set up a serverless SQL warehouse and a compute cluster with similar configurations. Run the "optimize table" command on the same table in both environments and measure the execution time. Repeat the test multiple times to get an average time. Finally, compare the average execution times to determine which is faster. Remember, actual performance can vary based on factors like specific configuration, resources, table size and complexity, and current system load and network conditions.

Good reads:

1. https://docs.databricks.com/en/admin/sql/warehouse-types.html

2. https://docs.databricks.com/en/lakehouse-architecture/performance-efficiency/best-practices.html

Please let us know if you have any further queries or concerns.

Best regards,

Yesh