[EPIC] Create a JH analysis service that recommends resource tiers

aicoe-aiops / operate-first-jupyterhub-analysis

Operate first Jupyterhub application provides a Jupyter notebook environment for python development and running machine learning workloads. In this repository, we analyze the data generated from the application hosted on the Operate First cluster on Openshift.

Other

2 stars 3 forks source link

[EPIC] Create a JH analysis service that recommends resource tiers #35

Open Shreyanand opened 2 years ago

Shreyanand commented 2 years ago

To make resource recommendation more easily accessible, we should create an AIOps service for it.

The service should have following components from a technical standpoint:

[ ] Given a cluster URL with Jupyterhub and Prometheus instance, fetch relevant metrics for the desired frequency
[ ] Run model training
[ ] Output recommended resource allocation policy for t-shirt size
[ ] Optional: Output user specific recommendation by inferring from trained model