On Pod18, these two tools are ready (https://wiki.opnfv.org/pages/viewpage.action?pageId=13208158)

1. Prometheus  (Expose prometheus, and access via 10.10.180.20:9090)
2. Litmus  (access vis 10.10.180.21:21:31826)
3. Grafana (access via 10.10.180.21:3000)

ToDo: Start Experiments 

Failure Data: Logs

Metrics from prometheus.

Nginx pods – Run Traffic. 

L3 Forwarder  - Run Traffic.

Log-Analysis :  Anomaly Detection

https://docs.google.com/presentation/d/1uWYkUJoeM4MFN-T2LRpEbPHT4PaCyXHHm_u8enkoPDg/edit?usp=sharing

All the updates regarding theoretical and experimental research will be put in this PPT.

Missing data in Logs is affecting the understanding of the problem. 

Lack of understanding of the dataset. Bert Implementation: (Not) available (https://github.com/google-research/bert).

Steve: https://arundo-adtk.readthedocs-hosted.com/en/stable/ : Good implementation to both understand and compare against.

No update on CFP.

Dataset will be used from above experiments.

Presentation was done to LFAI-Data.

Q: When you use BERT for log extraction, do you find any challenges in that data in the logs is not always actual words, sometimes short tokens?

The tokenization process of BERT.

Q; Failures are few and far (Physical and Virutal) - So, Failure Prediction problems in NFV world is not relevant.

Steve: Need to know more on this comment. Failure do occur, even with physical components.

Sridhar to follow up with LFAI-data.

Overview:

1. Used to provide dataset for industry to use.
2. Evaluation System - Evaluate the model (score). Fully Automated.
3. More Relevant to Intelligent-Networking scenarios
   1. Traffic Engineering.
   2. Capacity Planning.
   3. Security (network attack)

Present this next week?

New proposal to ITU - for CM-INP as platform for assessment