Anuket Project

2021-07-30 AI/ML for NFV Meeting Minutes


Attendees

Sridhar Rao

Rohit Singh Rathaur

Girish

Kanak Raj

Al Morton

Akanksha Singh


Sl. No.TopicPresenterNotes
1Committers and Contributors

Girish and Rohit will be Committers.

Rest will be Contributors.

Sridhar to complete the info.yaml - and accordingly create (a) Slack.

2Data Status

Failure Prediction using AI/ML in NFV Environments : The Data description is updated.

ToDo: (a) Location of the Node Failure events (b)  Location of the Link Failure events.

Failure Type           Data Model Status             Availability           Creation Possibility               
NodeIN PROGRESSNODifficult
LinksIN PROGRESSNOOK
VMYESYESDifficult (Experimental WIP)
ContainerYESNODifficult (Experimental WIP)
ApplicationIN PROGRESSYESDifficult
MiddlewareIN PROGRESSNODifficult
3Failure Prediction Implementation Status

2 Approaches to Enhance the model.

  1. LSTM - RNN. (ongoing) 
  2. GNN  - 

15th – Plan to complete both enhancements.

Complete the testing with existing data by End of August.

Implementation is Ongoing - Testing will be blocked until we get some data. Continuing with the existing data.

Jupyter-Notebooks.

If Testing of the models with more Real-world data is not possible - consider doing Gap-Analysis on (Trend/Pattern and Anomaly Detection)

3Volunteers - AlgoSelector

Kanak: Reinforcement Learning. Going through the articles. Unable to really find right reasoning. When to use which algorithm.

List the algorithms that are part of reinforcement learning - Start 2-3 algorithms. Applicability of these algorithms - when to use points.

Algo-Selector 

Implementation Approach:

Chatbot

Use: Decision Tree (Rule-Based). 

Goal: Which Python module/project to reuse.

4TVLV-Tool - Update.

APIs to config and start is completed. Stop API is ongoing.

ToDo: Containerizing and creating VM.

5