Azure Synapse versus databricks SQL endpoint performance comparison

[prasadvaze]

Has anyone done this and share details? I have a sample sql which ran on large SQL endpoint in 8min and synapse 1000DWU setting in 1hr. On small SQL endpoint it took 34min. What's the equivalent SQL Endpoint compute for Synapse@1000DWU? I know there is no 1:1 equivalent between DWU and DBU but what's the closest parity i should set to measure this sql performance

[BilalAslamDbrx]

@prasad vaze​, We don't share all benchmarks publicly, but I will say that Databricks SQL is very performant on a wide variety of workloads compared to Azure Synapse. For example, the Photon execution engine and Delta Cache are only available in Databricks SQL. Also, consider the overall architecture of your lakehouse -- it's just that much easier and simpler to build, manage and maintain on Databricks.

[prasadvaze]

What's the equivalent DBU for synapse 1000DWU?​

[Kaniz]

Hi @prasad vaze​ , This document contains recommendations on choosing the ideal number of data warehouse units (DWUs) for a dedicated SQL pool (formerly SQL DW) to optimize price and performance and how to change the number of units.

[Kaniz]

Hi @prasad vaze​  , Just a friendly follow-up. Do you still need help, or @Bilal Aslam​'s response helped you to find the solution? Please let us know.

[Anonymous]

Hey @prasad vaze​ 

Just wanted to check in if you were able to resolve your issue. If yes, would you be happy to mark an answer as best? If not, please tell us so we can help you.

Thanks!

[gfree76]

This is a difficult question to answer, because they are very different compute types. A better comparison would be the Azure Synapse Serverless SQL endpoints and the Databricks SQL. In that comparison, the databricks SQL endpoint is much much more performant, but also costs about 3x what the Synapse Serverless SQL compute costs. The costs for both types of serverless compute are still much lower than keeping dedicated compute resources running, whether Dedicated SQL Pool or Databricks interactive cluster.

[ABHI_CT]

The closest parity would be about 400 DBU on 8 core. the assumption is of same source and same SQL query in notebook. Also, the DC region has an iota of contribution to the speed. An assumption also includes the cost is out of the equation.

[prasadvaze]

The cost matters the most. If 400DBU on 8 core costs 3 times more than 1000DWU synapse compute ( assuming both are equal compute power) then why would one use the earlier? The comparison has to be on price-performance

[Lad]

Using AtScale to improve the scale, cost and concurrency of Databricks makes DB SQL shine over Synapse. ​

[Creddie]

I have use both, and as you said, there is no apples to apples match, but using the same workload, data bricks seems pretty fast compared to synapse, from my perspective.​

[Tomz]

One of our clients (an insurance company) recently decided to switch out the data warehouse platform from Azure Synaose to Databricks SQL based on cost, concurrency reason

[sammyadam]

DC region has an iota of contribution to the speed. An assumption also includes the cost is out of the equation.