Issues with data model compilation with worker threads

[igorlukanin]

Recently, we've improved the data model compilation by moving critical parts of the code to worker threads.

It can be enabled by upgrading to v1.2.13+ and setting the CUBEJS_TRANSPILATION_WORKER_THREADS environment variable to true. See #9188 for details.

If you encounter any issues or see any changes/improvements to your deployments, please report here.

[marianore-muttdata]

Hey there!
Just wondering, if we have something like 20 data models and 100 tenants, should we be able to see a different at all in request delays?
Do you have insights on the performance difference in scenarios with "smaller" data models?

[igorlukanin]

The smaller your data model is, the smaller the effect of this change is as well. Also, if your data model is compiled just once and not recompiled, then you'd to notice any difference at all.

The only case when the improvements should be evident is:

1. You have a large data model
2. It is frequently (re)compiled