AWS re:Invent 2025 - Build and optimize edge architecture for resiliency with AI (HMC403)

Building and Optimizing Edge Architecture for Resiliency with AI

Introduction to Hybrid and Edge Services

• AWS offers a range of hybrid and edge services to extend cloud capabilities closer to where data is produced and consumed:
  • AWS Local Zones: Extend AWS services to metro areas without a full AWS Region
  • AWS Outposts: Bring AWS compute and storage to customer premises
  • EKS Hybrid: Connect on-premises or virtual machines to a remote Kubernetes control plane
  • Edge devices: Services to support edge computing use cases
• These services enable low-latency, data residency, and other use cases for industries like telecommunications, media/entertainment, and gaming.

Defining Resiliency for Hybrid and Edge Environments

• Resiliency is the ability of a workload to recover from infrastructure or service disruptions.
• Key aspects of resiliency:
  1. High availability: Resistance to common failures through primary site design
  2. Disaster recovery: Backup and managed recovery objectives
  3. Operational resilience: Holistic resilience across compute, networking, data, and operations

Leveraging AI for Resilient Edge Architectures

• Idea: Use generative AI and agentic AI to leverage AWS expertise and design resilient hybrid/edge architectures.
• Approach:
  1. Gather data about the existing environment (e.g. outposts, local zones, availability zones)
  2. Leverage AWS documentation and best practices as knowledge sources
  3. Use an agentic loop with logical reasoning (e.g. via large language models) to analyze the environment and make prescriptive recommendations

Demonstration: Resilient Edge Architecture Assessment

1. Static Discovery: Use a Python script to discover existing edge infrastructure (availability zones, local zones, outposts, instance types).
2. Static Report Generation: Leverage Kira CLI to generate an HTML report summarizing the discovered edge resources and providing initial resilience recommendations.
3. Agentic Resilience Assessment:
   • Create a Strand agent to analyze the environment and AWS best practices.
   • Agent uses AWS knowledge base and tools (e.g. get outposts, get local zones) to gather comprehensive data.
   • Agent generates a detailed Markdown report with resilience analysis and actionable recommendations.

Advanced Agentic Patterns

• Agent as a Tool: Expose an agent as a reusable tool that can be consumed by other agents.
• Agent Swarm: Coordinate a group of specialized agents to collaboratively solve a complex problem.
• Agent Workflow: Define a structured workflow with transitions between different agent responsibilities.

Key Takeaways

• Hybrid and edge services enable low-latency, data residency, and other use cases, but require careful consideration of resiliency.
• Generative AI and agentic AI can be leveraged to automate the analysis of hybrid/edge environments and generate prescriptive resilience recommendations.
• Kira CLI and Strand agents provide a powerful platform for building resilient edge architectures using AI-driven assessments and recommendations.
• Advanced agentic patterns like agent-as-a-tool and agent swarms can further enhance the flexibility and capabilities of these AI-driven solutions.