Keynote Panel Discussion: Unifying the Edge

### Overview At the Open Networking Summit (ONS), a keynote panel explored the evolving landscape of edge computing, addressing the fragmentation, terminology confusion, and technical challenges across telco, enterprise, and IoT environments. The discussion centered on how the Linux Foundation Edge (LF Edge) initiative aims to unify these disparate domains through open-source collaboration, vendor-neutral governance, and standardized architectural layers.

### Key Topics - **Defining the Edge**: Moving compute as close as necessary and feasible to subscribers or devices, ranging from cloud edges and telco CPE to deep industrial/embedded environments. - **Fragmentation vs. Open Source**: Historical proprietary lock-in models versus the maker movement and open standards where interoperability scales across multi-cloud and edge deployments. - **LF Edge Projects**: The roles of key community efforts including EdgeX Foundry, Project EVE, Akraino, Baetyl, Fledge, Home Edge, and the Edge Computing Glossary. - **Workload Continuum and Architecture**: Balancing centralized deep learning in the cloud with localized shallow learning, containerization, virtualization, and physical constraints at the edge.

### Technical Details - **Edge Heterogeneity and Hardware Constraints**: As compute moves toward the device edge, hardware requirements diversify to include ruggedization, wide temperature ranges, diverse I/O, explosion-proof ratings (Class 1 Div 2), and memory limits where containerization and virtualization cannot operate (the thin compute edge). - **Decoupling Data and Southbound Protocols**: EdgeX Foundry addresses thousands of southbound protocols by introducing an open API and loosely coupled microservices architecture. This decouples data collection at inception, enabling deployment across any edge-to-cloud topology without locking users into specific hardware, OS, or cloud backends. - **Security and Root of Trust**: In deep enterprise and industrial edges (e.g., servers mounted in ceilings or beside generators), physical access is limited. Implementing hardware root of trust, measured boot (via Linux or RTOS), and virtualization layers (such as Project EVE) abstracts underlying physical interfaces (like serial or proprietary radios) into standard Ethernet-like connectivity for cloud-native applications. - **Akraino and Validated Stacks**: To avoid the fragmentation previously seen in cloud deployments (e.g., mismatched OpenStack and Kubernetes configurations), Akraino focuses on delivering fully declarative, validated blueprint stacks for specific use cases across integrated hardware and software. - **Terminology Precision**: The panel emphasized replacing subjective terms like "near/far edge" and vague references to "real-time" or "tiny" with absolute metrics (e.g., sub-5 millisecond 5G latency, deterministic microsecond responses for critical systems like airbags, specific memory footprints).

### Takeaways Edge infrastructure is expanding rapidly, with vast amounts of operational technology (OT) and industrial infrastructure already physically in place but disconnected. 5G cannot replace edge compute due to bandwidth economics, physical latency limits, and safety-critical deterministic requirements. Successful edge scaling relies on open-source, vendor-neutral infrastructure to avoid "reinventing the middle," enabling organizations to focus on specialized domain applications and algorithms.