...
Antitrust policies
- Linux Foundation Anti-Trust Policy
- GSMA Anti-Trust Policy Notice
- Recorded Policies:
...
Thoth: AI/ML for NFV Usecases (An Incubation Project) - Sridhar Rao
Domain: Failure Prediction. (for works in this domain, and other domains, please refer to the survey).
Failures of: VMs, Containers/Pods, Nodes, Application* (access), Network-Services.
For each of these failure - a separate model is required.
Existing works - Only VM failures are studied very well. All these works consider Infrastructure data.
Data Required:
* Time Series
* Failure Events
* Infrastructure data from 'n'-Days before the failure occured till the time the failure happened. Infrastructure Data: CPU, Memory, Interfaces, Storage, H/W, VNF-specific resource consumption.
How failure is defined:
- Nodes - Shutdown/reset
- VM - Shutdown/reset
- Containers - Shutdown/reset.
- Application - Access.
- Network-Services - Access.
Access to the Data:
- Current sources - https://docs.google.com/spreadsheets/d/1QgxlPj8siTLc0ZAggPf1l-GoATqqqOij3GiracwQ3oQ/edit#gid=0
- Collaboration - EUAG, OpenInfraLabs - Telemetry WG, Other Researchers.
- Generate from Testbed - Academic Openstack Testbed, Kubernetes Kuberef-RI2 Testbed (pod18 Intel) – Chaos Engineering + Barometer Collectd - Create Data.
- Generate Synthetic Data - GANs.
Ongoing Efforts
- Enlist the operations that are done from VM/Container, and that can make the VM/Container fail – this is to emulate failure event. Time-based, configuration-changing, stress-ng (supports multiple Dimensions).
- AlgoSelector – Series of Qs that are asked to user about the data and the problem - The tool will suggest the Algorithm to use in in one of (Supervised, Unsupervised, Reinforcement).
Artifacts:
- Models - Framework Independent (Jupyter Notebooks), Framework Specific - python files
- Tools - Python Files.
- Dataset - Kaggle
Network automation project - Sridhar Rao , Jie Niu
...