Analytics for automating critical infrastructures
0 likes97 views
This document discusses using network analytics and machine learning to automate critical infrastructures. It outlines a path towards network automation using simplified service ordering, streamlined operations, and intuitive interfaces. Analytics can provide insights to optimize network performance through continuous data collection, analysis using AI models, and automated orchestration. Open challenges include high data requirements, multi-vendor data sharing, and integrating ML solutions into network operations. Case studies demonstrate ML-based optical network monitoring and a data-sovereign telemetry broker. Overall, network analytics and AI will enhance optical network control and automation.
![2023 © ADTRAN, INC.
7
• Machine Learning solution for OTDR traces
• Reflective event detection in Passive Optical Networks (PON)
• Web-UI for visualization of PON characterization and monitoring
UC1 ML-Based PON Characterization
Research highlights
e d e e
n e
e
e
n
e
onne o o e e
on o
e
o b ne
Maximilian Brügge et al., “Live Demonstration of ML-Based PON
e on nd on o ng ( Z.7)”, OFC 2023 [M3Z.7]
2023
M3Z.7](https://image.slidesharecdn.com/230424-ofc23-w4a-230424141210-d9fc6e49/85/Analytics-for-automating-critical-infrastructures-7-320.jpg)
![2023 © ADTRAN, INC.
8
• Modular telemetry broker for extensive collection and sharing of data
• Flexible Optical Terminal in a partially disaggregated optical network
• Optical performance measurement (SNR, Q-Factor, BER) with second granularity
Research highlights
UC2 Data-Sovereign Telemetry Broker
`
CA
ROADM 3
ROADM 1 ROADM 2
OLS
Vendor B
λ2
Terminal
Vendor
C
10G/100G
aggregation
10G/100G
aggregation
λ2
Terminal
Vendor
A
agent
agent
CB CC
Data Sovereign
Marketplace
Network Health
Monitoring Tool
IDS Connector
VNF Provider Y
Impairment
Validation Tool
IDS Connector
VNF Provider Y
Data Marketplace Connector
Data MarketplaceConnector
Mapping Data
Model
DSC
Agent Wrapper
Consumer
DSC
Data
Owner
Config File
Agent Wrapper Provider DSC Consumer DSC
Agent Wrapper Provider DSC Consumer DSC
Collect Device Info
Create Resources
Artifact Request
Call / Telemetry
Device Info
TelemetryData Res. Serialize Telemetry
Data
Artifact Response
Payload: Serialized Data
Telemetry
Update
Telemetry
Req.
TelemetryReq.
2023
M3Z.3
Haydar Qarawlus et al., “Demonstration of Data-Sovereign Telemetry Broker
o en nd gg eg ed e o k ”, OFC 2023 [M3Z.7]](https://image.slidesharecdn.com/230424-ofc23-w4a-230424141210-d9fc6e49/85/Analytics-for-automating-critical-infrastructures-8-320.jpg)
Analytics for automating critical infrastructures
- 1. Analytics for automating critical infrastructures Achim Autenrieth, Adtran OFC 2023, 8 Mar 2023 Symposium on “beyond the hype of network analytics: use cases, feasibility, and barriers.” ADVA is now part of Adtran
- 2. 2023 © ADTRAN, INC. 2 The path towards network automation Simplified point-and-click service ordering Speed-up Service Monetization Streamline Network Operations Simplify Support Automated, fully interconnected, virtualized network infrastructure Intuitive customer-facing interfaces that simplify service management and monitoring Intent-based operation Streaming telemetry & AI-based analytics AI-based network automation SDN control ✓
- 3. 2023 © ADTRAN, INC. 3 AI/ML for network automation and optimization Programmable Network Insights Orchestration Open APIs Streaming Telemetry Closed Loop Automation Data collection (Telemetry) Analysis (Real-time & offline) Refine & Adapt automatically Predictive Maintenance Turning data into actionable insights to optimize network and service performance Network Optimization & Insights Turning data into actionable insights to optimize network and service performance • Focus on end-to-end optimization • Combines real-time data collection, AI-driven analysis and orchestration to enable proactive optimization • Cloud-based control covering on-premise, access and packet/optical metro/core networks Streaming telemetry • Continuous data collection via streaming telemetry • Data-centric analysis of network conditions • gRPC / gNMI and NETCONF/YANG protocols Network analytics / ML use cases • ML-based transmission performance optimization • Traffic and failure prediction • Predictive maintenance • ML-based network optimization
- 4. 2023 © ADTRAN, INC. 4 Network analytics pipeline Data Storage and Analysis Time series database Visualization Data Collection Data Buffering Stream processing Message broker • gRPC/gNMI • NETCONF • SNMP • YANG Push Telemetry Network Elements ML-assisted Solutions Ticketing Alarming SDN Control Orchestration Scalable common data collection framework • Telemetry retrieval • Efficient telemetry collectors, brokers, and Time-Series Databases (TSDB) • Computation requirements for data analysis • Cross-interface and cross-terminal telemetry data sharing
- 5. 2023 © ADTRAN, INC. 5 AI-driven fiber networking – the whole picture Kafka (streaming telemetry/events) RESTCONF (prov and monitoring with service level APIs) Cloud-based network analytics applications AI-Driven Orchestration, Management and Optimization Network and service control and orchestration Network Optimization & Insights NETCONF/YANG control gRPC telemetry OSS / BSS APIs (integration on op. / business support systems) uCPE Edge cloud Cloud-managed Mesh Wi-Fi Business Ethernet Fiber Access Ethernet/WDM with network synchronization DCI Mobile X-Haul and Wholesale Metro / Core WDM Access and Aggregation Optical Networking Subscriber Networks / Solutions Open, disaggregated access, aggregation and transport Cloud data center Core data center Metro data center Open Line System (OLS)
- 6. 2023 © ADTRAN, INC. 6 Open questions & challenges Can we extract real-time data for ML models? Can we work with the data we already have? Can this be orchestrated on an SDN-based network? Can we address relevant use- cases to our ML+data setup? Can we share data with data privacy and sovereignity? How do we get the data to the orchestrator ? UC1 UC2
- 7. 2023 © ADTRAN, INC. 7 • Machine Learning solution for OTDR traces • Reflective event detection in Passive Optical Networks (PON) • Web-UI for visualization of PON characterization and monitoring UC1 ML-Based PON Characterization Research highlights e d e e n e e e n e onne o o e e on o e o b ne Maximilian Brügge et al., “Live Demonstration of ML-Based PON e on nd on o ng ( Z.7)”, OFC 2023 [M3Z.7] 2023 M3Z.7
- 8. 2023 © ADTRAN, INC. 8 • Modular telemetry broker for extensive collection and sharing of data • Flexible Optical Terminal in a partially disaggregated optical network • Optical performance measurement (SNR, Q-Factor, BER) with second granularity Research highlights UC2 Data-Sovereign Telemetry Broker ` CA ROADM 3 ROADM 1 ROADM 2 OLS Vendor B λ2 Terminal Vendor C 10G/100G aggregation 10G/100G aggregation λ2 Terminal Vendor A agent agent CB CC Data Sovereign Marketplace Network Health Monitoring Tool IDS Connector VNF Provider Y Impairment Validation Tool IDS Connector VNF Provider Y Data Marketplace Connector Data MarketplaceConnector Mapping Data Model DSC Agent Wrapper Consumer DSC Data Owner Config File Agent Wrapper Provider DSC Consumer DSC Agent Wrapper Provider DSC Consumer DSC Collect Device Info Create Resources Artifact Request Call / Telemetry Device Info TelemetryData Res. Serialize Telemetry Data Artifact Response Payload: Serialized Data Telemetry Update Telemetry Req. TelemetryReq. 2023 M3Z.3 Haydar Qarawlus et al., “Demonstration of Data-Sovereign Telemetry Broker o en nd gg eg ed e o k ”, OFC 2023 [M3Z.7]
- 9. 2023 © ADTRAN, INC. 9 • SDN, streaming telemetry, and network analytics provides you insights to improve and optimize network operations • Privacy-preserving ML techniques and inter-operator model sharing support carrier grade network automation • AI/ML will gradually enhance / augment optical network control and automation Open research challenges & next steps • High data acquisition and processing requirements • Alignment on multi-vendor ML-AI data exchange formats • Open dataset access and machine-learning marketplace integration • Transport network simulation and digital twin • Network optimization, energy efficiency and sustainability • Integration in vendor solutions and interface to OSS / BSS Conclusion and outlook
- 10. Thank you This work has been performed in the framework of the CELTIC-NEXT project AI-NET-PROTECT (Project ID C2019/3-4), and it is partly funded by the German Federal Ministry of Education and Research (FKZ16KIS1279K). Acknowledgements: Maximilian Brügge, Jasper Müller, Sai Kireet Patri, Sander Jansen, Jim Zou, Stephanie Althoff, Klaus-Tycho Förster, Haydar Qarawlus, Steffen Biehs, Behnam Shariati, Jose-Juan Pedreno-Manresa, Ayoub Bouchedoub, Hendrik Haße, Pooyan Safari, Johannes Karl Fischer