What is the difference between those 2 products (Cumulocity IoT Machine Learning/ML Workbench and Zementis)
Cumulocity IoT’s machine learning capabilities are powered by Zementis behind the scenes.
Machine Learning Workbench is a new tool that allows you to build and train models directly using Cumulocity IoT (whereas Zementis is predominantly an execution environment for pre-built models).
There are more introductory notes on the new MLW capabilities here:
There is a little bit of a history here… Zementis was the company that software AG acquired. We incorporated their standalone machine learning runtime into the Cloud as a Cumulocity IoT ML micro-service. In the latest 10.7 release, we have extended the runtime (or inference engine) with an ability to also train models.
Ok Thanks for the clarifications. We want to explore Zementis and test the ML features (pre-process data, build and deploy model, etc). Should we run Zementis server seperately or just use the integrated micro-service in Cumulocity ? What are the pros/cons of each option ?
We also have, Cumulocity IoT edge instance, is it easy to have Zementis or Cumulocity IoT ML installed on the edge ?
The Cumulocity IoT Machine Learning would be the recommended way forward as it is available as a microservice on the cloud AND on the edge server. Subscribing to this will automatically provision a zementis (inference engine) micro-service for you. Depending on the use case, you would also have the flexibility to choose additional micro-services e.g. use ONNX if you are using deep learning (image recognition), NYOKA for time series analysis, MLW for model training, pre-processing data, 1-click deployment etc.
The Zementis server is more of a standalone server that enables machine learning integration with other products so it would not be relevant for you.