Azure Fabric warehouse performance

[JP_Helsinki]

Hi!

We have been testing Azure Fabric warehouse for our internal datamodel purposes (currently inhouse SQL) and we are using Knime ETL to load calculated data into the warehouse. The connection uses the endpoint connection to the DW (xxx.datawarehouse.pbidedicated.windows.net:1433)

Im qurious about the insert times that we are facing with the Fabric Warehouse. A simple insert statement with 10 columns and 5000 rows takes about 5-7 minutes to run, and insert of 80k rows takes hours. I have tested this with azure serverless SQL which is alot faster (still slow compared to inhouse SQL) but the Fabric performance seems unusable. 

We have the lowest capasity (F2) but when looking at the Fabric Capacity Metrics the CPU at least does not seem to be the bottleneck here. 

Any ideas how to boost the performance? 

[JP_Helsinki]

Thanks Josh, pretty bad is an understatement in my opinion 🙈 just tested the same dataset in a local environment and it takes less the a second  to complete. Is this due to the capacity limitations (more expensive would be faster) or?

[Joshrodgers123]

I think this is unfortunately just the current state of the platform. It's supposedly GA, but so many components of Fabric are either half-baked or broken altogether.

The "unlimited" free trial seems to be coming to end. I was hoping they'd get things together before that happened. 

[Joshrodgers123]

Yeah, I agree it's pretty bad. I'm trying to use DFG2 to load data into a warehouse. I'm seeing really small tables with only a couple hundred rows take 10 minutes to write when they decide to work. Most of the time they just time out and fail.