1. Closed Loop Platform Automation w/ OPNFV
Sunku Ranganath, John Browne, Tim Verrall, Damien Power, Emma Collins, Krzysztof Kepka
Thanks to Reviewers: Trevor Cooper, Jack Morgan, Matthias Runge, Tomi Juvonen, Tomofumi Hayashi, Toshiaki Takahashi, Yuki Kasuya

2. Agenda Introduce Closed Loop Automation (CPA) & OPNFV intercept
Propose CLA Working Group
Source: +WG+Proposal

3. Real-time/Near Real-time Loops - Automated
Multiple Closed Loops
Use cases (Loops)
Capacity planning
Peering planning
Cache placement …
Plan & Provision
Telemetry
Offline Processing
Offline
feedback loop
Forecast
Optimize
Design
Control
Analyze
Telemetry
Assure
Near-real
Time
Feedback loop
Use cases (Loops)
Service assurance
Security operations …
Online Processing
Use cases (Loops)
Traffic Engineering: Network Optimization
Demand placement
Workload placement…
Real-Time
Feedback loop
Telemetry
Orchestrate
Monitor
Real-time/Near Real-time Loops - Automated
Source:

4. Networking Closed Loops – High Level Architecture
Policy Based Provisioning
Analytics
Systems
Business Applications
Setting of Policy
Control Loops
Application
Telemetry distribution or storage or …..
Cloud and Virtual Management
MANO
EMS
VNFM
Control
SDN/NMS
Network Services
Platform Telemetry agents: Collectd -
Local Platform Agent: Resource Management Daemon -
Platform
Infrastructure
Platform
Telemetry
Local
Platform
Agent
Forwarding Plane
Interfaces
Interfaces
Platform Resources
Traffic
Traffic
Independent Closed Loops: SDN, Cloud & Virtual Mgt, Platform

5. Closed Loops – Use Cases
Policy Based Provisioning
Control Loops
Security
Improved Customer Experience
Cloud Optimization & Efficiency
Business
Use Cases
Threat Detection
Edge Placement
Energy Optimization ?? ??
Threat Response
Service Healing
OPNFV Doctor
Capacity Optimization
OPNFV NSB
OPNFV Bottlenecks
Service Optimization
Cloud Configurations ??
Differentiated QoS
Business Applications
NFV Orchestrator (NFVO) [eg ONAP/OSM]
Analytics
AI/ML/DL
VNF Manager (VNFM)
OPNFV OVP
Question about use cases of interest. Looking to help these with IA features. Feature exposure, provisioning & telemetry. We are looking to enable/fill in these gaps.
Telemetry I/F
Open Stack
Bare Metal
Kubernetes
Telemetry I/F
Monitoring/Storage
Telemetry I/F
Scope of Interest
Feature Exposure
Provisioning
Telemetry
OPNFV Barometer
Local Policy Enforcement Agent(s)
collectd
For Local Dynamic Control
Platform(s)
Intel Infrastructure Management Tech
Intel RDT
Power
Intel RunSure

6. Use Case List Noisy Neighbor avoidance
Energy efficiency with improved IDLE power consumption, power management, etc.
Power aware workload placement
Reliability aware workload placement
Reliability aware auto-scaling
And so on…
Note: This is not an exhaustive list

7. CLA WG Proposal Objectives: Scope
Common platform to discuss, solve, prioritize, implement CLA architectures, usecases, & solutions
Drive upstream adoption of closed loop usecases across NFVI
Support cloud native, appliance and virtualized infrastructure models with offline/real-time closed loops
Scope
Forum across all OPNFV projects to facilitate enhancements, APIs, methodologies, testing
Forum to discuss and inter-network among projects to facilitate feasibility among projects
Bridge to communities under LF – LFN, LF Edge, ONAP, etc., on closed loop automation

8. Open Source Communities - Possible Intercept
Communities of Interest
Telemetry Projects – CNCF OpenMetrics, Collectd, Telegraf
OpenStack Self-Healing SIG, Ceilometer, AODH, Mistral, Congress, Vitrage, Fenix, Monasca, etc.
Kubernetes “telemetry based scheduling”
ONAP CLAMP, DCAE
Akraino Closed Loop Blueprints
O-RAN closed loop use cases
PNDA
Note: This is not an exhaustive list. Provides possible intercept with various open source communities of interest

9. Work Items Identify closed loop use cases of interest
Establish strong connection with ONAP & LFN communities to achieve complimentary solutions/implementation
Identify gaps necessary to implement CLA use cases of interest
Produce reference solutions/implementations for initial set of use cases leveraging existing software capabilities & tool-sets within participating OPNFV projects
Drive medium-long term changes in collaboration with orchestrators like OpenStack & Kubernetes communities to  achieve closed loop implementations

10. Demo #1 Closed Loop Resiliency demo with OpenStack Ongoing discussion
Participating projects: Doctor + Barometer
Provides closed loop resiliency and maintenance with Vitrage+Congress+Fenix
Produce demo, reference code and documentation
Unique contributions: Additions to Vitrage + Fenix API for closed loop implementations

11. Backup

12. Telemetry available through Collectd

13. Networking Closed Loop Groups Real-time/Near Real-time Closed Loops
Service Optimization
Service Healing
Threat Detection
Differentiated QoS
Capacity Optimization
Edge Placement
(Closer to the edge)
Threat Response
Energy optimization*
Energy optimization*
End to End Assurance
Network Routing/
Security
[Routing]
Automation
Orchestration/
Placement/
Provisioning
Automation
Auto-scaling
Operations
Reliability
End to End Assurance
Performance Management
[Equipment]
Management and Analytics
Telemetry
Actions
Platform
Intel Run Sure/RAS
Intel RDT
Power
Groups of Control Loops

14. Collaborate to Integrate w/ OPNFV Projects
Intercept – Existing or New?: Few projects already leverage parts of CLPA concepts
OPNFV Doctor -> self-healing & fault management using OpenStack Fenix (1)
OPNFV Bottleneck -> to implement AI over test data (2)
OPNFV VSPERF -> to implement analytics over performance results
OPNFV NSB -> sample VNFs integration with platform metrics
OPNFV Barometer -> telemetry
Unify the CLPA approach
Provide uniform approach to closed loop automation
Enable newer use cases leveraging advanced platform telemetry in energy efficiency, differentiated QoS, security, etc.
Get out telemetry story. 1. 2.