边缘计算演进_17页_4mb

2 Why Evolve Edge Computing?

Arm calls for an evolution in edge computing due to convergence of several trends: increasing connectivity between edge devices and the cloud, the massive shift toward edge AI for insights generation, and growing security and management requirements at scale. This evolution requires a 'cloud-like' approach, but adapted to edge constraints like hardware heterogeneity and limited connectivity.

Evolving Edge Computing – Essential Ingredients

To achieve this evolution, four essential ingredients were identified:

1. Cloud-like Agility: High-level development offering frictionless development with hardware abstraction and full hardware benefits exploitation.
2. Security at Scale: Trusted devices with secure software lifecycle management and regulatory compliance.
3. Hardware Efficiency: Optimized computing for specific use cases.
4. Collaborative Ecosystem: Eliminating unnecessary fragmentation and promoting modular, interoperable software.

Vision for Evolving Edge Computing

The vision is frictionless 'cloud-like' development on Arm architecture, fueled by collaboration, maximizing re-use, accelerating time-to-market, and reducing total cost of ownership (TCO). This especially involves adapting cloud-native principles like CI/CD.

Key Changes Driving Edge Evolution

• Ever-increasing scale and complexity of connected edge deployments, demanding sophisticated data insights. However, constraints like latency, power, cost, privacy, and connectivity require processing close to the data source.
• Security at Scale: Growing regulations and inherent risks of edge deployments (susceptible to attack, harsh environments) demand a robust, consistent trust model (secure identity, secure updates, etc.).
• Operational Efficiency: Long device lifetimes (5-10 years+) make TCO a major factor, focusing on power consumption, maintenance (secure updates), and development costs.
• Agile Innovation: The need for faster time-to-market and continuous improvement requires the adoption of agile development flows (CI/CD) similar to cloud.

Challenges to Overcome

1. Developing a 'Cloud-Like' Mindset at the Edge: Moving away from traditional 'write-once-run-anywhere' embedded approaches. Requires adopting agility (CI/CD) and common abstracted programming models for portable development across diverse edge hardware.
2. Security and Privacy at Scale: Ensuring trusted, consistent security for complex multi-vendor software stacks and diverse threat environments. Requires composable software with private, secure lifecycles, platform security capabilities (secure boot, etc.) from the outset.
3. Eliminating Needless Fragmentation: Reducing fragmentation for innovation and cost efficiency. Requires defining standard hardware/software interfaces (like defined boot processes via Arm SystemReady) and common security capabilities.
4. Heterogeneity in Edge AI: Managing complex, data-intensive edge AI workloads across heterogeneous hardware (CPU, GPU, NPU) requires developer experiences that abstract hardware variations effectively and do not cause fragmentation.
5. Evolution of Industry Collaboration Models: Achieving the vision requires standardization and interoperability across complex multi-vendor stacks, similar to other industries (e.g., telecoms). This helps eliminate differentiation that adds no value, reducing TCO and enabling scale.

In summary, edge computing must evolve beyond current approaches by adopting a cloud-like operational model enhanced with agility, robust security, and collaborative development principles tailored to its unique constraints and scale requirements, particularly crucial as edge AI becomes a major driver of growth.