Forum Discussion

yan's avatar
yan
Advocate I
6 years ago

DirectQuery intentionally slow with non-MS sources ??

Catchy title but fact is that executing a simple query (select from small table) from SQL Server renders immediately but a similar one from Google BigQuery for example adds a lot of untraceable overhead.

 

To reproduce:

  1. download this tiny pbix - it has a single map visual on public data on bike stations in Austin. That is, 96 records.
  2. log in with your Google GCP account
  3. enable Performance analyzer and hit Refresh
  4. note as Direct query is reported to take > 4s
  5. in BigQuery's console, Query history shows the actual runtime was < 0.1s ...

Interestingly subsequent Refreshes take no time in BigQuery because result is cached but PBI still reports it took 4s .

 

Under the hood, I know BigQuery embedded connector relies on ODBC using a slim version of the official Simba driver but if ODBC is at fault, why has MS not used the native .Net library. That design choice is dubious as the DirectQuery experience could be way much better if ODBC is the bottleneck - hence the title.

Any suggestion is welcome - or maybe has someone written a .Net custom connector for BigQuery already, using its API instead of ODBC ?

 

Others using ODBC-based embedded connectors (e.g. Snowflake) could face the same issue so interested in feedback

1 Reply

  • Same here with the Vertica driver. The Simba version is not just slow, it is also out of date and broken. Thankfully we could install the genuine ODBC driver from the Vertica site instead (with the added benefit of custom query).

     

    What I find outrageous is that the Simba drivers do not show up in the ODBC control panel.