Use of MMS
Traditionally, using an ML-based inferencing service API running in the Cloud is an accepted approach. However, for reasons such as the need for low latency, data privacy, lower cost, and autonomous operation, the application may want to perform the task of ML model-based inferencing at the Edge.
That will require that the ML model be delivered to the Edge node where needed.
As an alternative solution, some applications may choose to wrap the ML model along with the application code itself within the same container and deploy that composite image to the Edge node. Given the life cycle of frequently updating the ML model, which is usually different from the application code, if you use the composite approach, every time the model is updated, you will need to transfer the entire composite application image to the Edge. This technique also has the disadvantage that your service will suffer downtime for a period while the old container is stopped and the updated container...
