Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

This project aims to build machine-Learning models and tools that can be used by Telcos (typically by the operations team in Telcos). Each of these models aims to solve single problem within a particular category.

Requirements


CategoryJira ReferenceDescription
Modelhttps://jira.anuket.io/browse/THOTH-1 
  1. Existing: Cleanup and publish the existing (FP) one. We may have to publish in a more usable form (part of a framework, or Python program).
  2. Ongoing: Log Analysis (BERT). Openstack Logs (nova-*, neutron-* ...). Currently openstack logs from a very small (3 computes, 2 controller, 1 Storage) openstack setup is being used.
Toolshttps://jira.anuket.io/browse/THOTH-2
  1. Existing: Cleanup the AlgoSelector.
  2. Ongoing: Data Selection/Anonymization.
  3. Ongoing: Chaos* (stress-ng)  – Used in the testbed (Intel Pod18, for ex.. but restricted to K8S cluster)

Framework

(MaaS)

https://jira.anuket.io/browse/THOTH-1 

Upstream: Kubeflow/Acumos.

Thoth: Kubeflow + Supervised Techniques + Integration with important Data-Pipelines/Sources

WhitePaper

(Research effort)

https://jira.anuket.io/browse/THOTH-3AI/ML & Kubernetes (CN-NFV)




Provide a list of features or use cases, documented as Epics or Stories in Jira.  Use the Jira issue insertion feature for Confluence.

...

NameDescription

Format (Container, Compressed File, etc.)

ML-Models
  1. Failure Prediction Model
  2. Log-Analysis
  1. Jupyter notebook
  2. Python Application (Adaptable to ML-Framework)
  3. Containerized ML-Model (Kubernetes based ML-Framework).
ToolsData ExtractionPython Application - Jupyter Notebook
Research Studies

AI/ML problems in NFV, OSS Frameworks

AI/ML & Kubernetes in NFV

.md files and/or pdf files.
ML-FrameworkUpstream ML framework projectIntegration Code.

Architecture

High level architecture diagram

...