Base Table Maximum threads to be setup per cube

[fschleich]

Hello,

Currently the number of queries running in parallel during an EC build is set at the server level. With setting "Build / Base Table Max Threads" (4 by default).

The optimised number of simultaneous queries is very much dependant on the context of the EC. For instance, an EC with 4 differents date source is more likely to accept a high number of simultaneous queries (spread over the 4 data sources), than an EC with only one data source.

Therefore, I would like to suggest to make this setting per cube instead of at the server level. An alternative would be to have this setting at the Data Group level.

Thank you and best regards,

Franck

[DRay]

Hello @fschleich.

Thank you for your feedback! I would like to get this Idea over to the product team, but first I would like to get more information and head off any roadblocks to implementation.

The issue I'm worried about is that there can be a lot of data models in Sisense, and configuring maximum threads per each could quickly overload the system resources, adding complexity and risk without any appreciable increase in performance. What are your thoughts?

Thank you for working with me to get this Idea fleshed out so it can be evaluated and scoped for effort.

Have a great day!

[fschleich]

Hi Dray,

Let me try to give you more details about eh option "Base tables Max Threads"

1. Current situation

The option "Base tables Max Threads" is set in the menu "System Configuration", hence the same for every cubes

2. My initial proposal :

Be able to define this value per cube. But I understand your concern about a risk of overload.

3. Alternate proposal #1 :

Be able to define it per Data Group. Having a new option "Base tables Max Threads" in this menu :

4. Alternate proposal #2 :

Be able to define it per Connector. Having a new option "Base tables Max Threads" in this menu :

=> My prefered alternate option is #2 (per connector) 

Indeed, in makes sens to define the parallelism level per type of data source (ie : when accessing a standalone database is not the same as accessing a clustered database where we can increase the number of queries run in parallel)

What do you think ?

Kind regards,

Franck

[DRay]

Hi @fschleich,

Thank you for sharing your idea and providing details. We will bring this to the product team and provide updates here as they are available.

We encourage people to continue to vote and comment. We use community interest as one of the ways we prioritize efforts, so every vote and comment can help turn a request into a reality.

Thank you again for sharing and being part of our community!