KNIME Server and Column Storage

[supersharp]

Hi there
How do you make the columnar backend the default for all workflows on KNIME AP. Is there a way to make it the default method for Server 14.4.1?
Thanks!

[spider]

Are there performance improvements on the KNIME server?

[nemad]

Hello @supersharp,

you can set the Columnar Backend as default for new workflows in AP 4.5.1 via File->Preferences->KNIME->Table Backend.
There is currently no way to migrate all workflows on the Server because there are slight differences between the backends.
The most notable of those differences are:

• No support for error causes. If a value is missing, then the backend doesn’t store any data for it. This most notably affects Extract Missing Value Cause – KNIME Hub.
• Upcasting of primitive values in collections. Example: If you put an integer into a collection with doubles, then it implicitly becomes a double and will not be turned back into an integer when the collection is split again.

We recommend to build new workflows with the Columnar Backend and migrate existing workflows in which you encounter performance issues. Note that the former backend has not been deprecated yet and even when it is deprecated, it will still stay a part of the AP to ensure backwards compatibility.

Regarding @spider 's question: That depends a bit on what you mean by performance on the KNIME server. If you run workflows via an executor on the server, then yes, the Columnar Backend will improve the performance of those workflows that make use of it, especially if they are dealing with large datasets.

Kind regards,
Adrian