Training request mechanics

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
05-17-2024 05:36 AM - edited 05-17-2024 05:37 AM
Happy Friday,
I'm trying to demonstrate PI to a customer base. While I can set it all up and demonstrate from my Personal Dev Instance, any attempts at training a model from their pre-prod are failing training.
I'd like to ask something a bit more detailed than what current training or reference materials have provided. When your instance attempts to train one of the "out of the box" Classification Definitions and you get a state result overnight of "Training request timed out", the first question that system and network admins say is "well, if the training service is within ServiceNow's data center, why would there be any firewall or network issue causing a timeout?"
I tried to explain that the training service logs into the instance web service via the sharedservice.worker user ID, but is there a clear document showing/describing the mechanics of that exchange for technical purposes?
1. Show the details of the API call from your instance to the training service. Ports, content etc
2. Show the details of the call from training service to your instance to retrieve the Classification Definition data from the ml_capability_defintion_classification table
3. Show the details of the call from training service to your instance to pass back the machine learning model artifacts, including passing the model to the ml_solution table record etc
Thank you in advance if such a technical artifact exists. Sadly, most all the training material and documentation glosses over this level of detail and just lists general suggestions on what to check if you get a timeout.