Hey folks,
I’m building SARIMA model for forecasting, while running on KNIME Analytics Platform it is taking 2 - 3 mins to execute. But same workflow deployed on KNIME Business Hub, it is taking more than 10 mins & often not able to run completely also.
Could you help me on this.
Hi @Umesh_0624 what do you see when it doesn’t run on Business Hub, can you show us the error?
How does the workflow look like? Even a screenshot would help us gain an overview.
Are the Analytics Platform and Business Hub Execution Context (Executor) on the same version?
On the Hub, are you executing it as an ad-hoc job or as a scheduled job or data app?
Hi @Dora_Oravecz this workflow is related to Manufacturing Demand Forecasting have downloaded from the community.
I’m running this as an ad-hoc job & takes longer duration >15 mins & still workflow is not completely executed (Refer attached image). While on desktop < 2 mins.
And Executor is of latest & same version.
NB - I think there is some issues on SARIMA nodes. As our other workflows are running smoothly.
Hi @Umesh_0624 , thank you for the details.
Please note that the SARIMA nodes are Labs, meaning they are experimental nodes that are not fully finalized yet.
Hence, the nodes work fine locally but might not perform as well on Business Hub as their use is not recommended in production.
Additionally, the workflow does predictions but it’s intended as a training workflow in the context of time series analysis. It’s not supposed to be deployed as is on Business Hub. For the deployment workflow, you’d save the best model and apply it to new data using just the Predictor node - you don’t train the SARIMA Learner every time.
Hi @Dora_Oravecz thank you for your suggestion / input.
