AWS Architecture Blog·September 1, 2026Hybrid Cloud Orchestration for On-Premises Infrastructure with AWS Serverless
This article details a hybrid cloud orchestration solution leveraging AWS serverless services (Lambda, Step Functions, DynamoDB, EventBridge) to manage distributed on-premises infrastructure, including bare-metal servers and EKS Anywhere clusters. It addresses challenges like inconsistency, manual bottlenecks, and fragmented visibility across geographically dispersed data centers by providing a centralized control plane in the cloud with distributed execution on-premises. The architecture emphasizes event-driven workflows and a centralized inventory management system for robust and scalable operations.
Read original on AWS Architecture BlogThe article presents a robust solution for hybrid cloud orchestration, focusing on managing extensive on-premises infrastructure from the AWS cloud. This is particularly relevant for organizations with large, geographically distributed data centers that require bare-metal configuration, deployment, and lifecycle management, often with Kubernetes control planes (EKS Anywhere) residing on-premises due to regulatory or connectivity constraints.
Key Challenges in Distributed On-Premises Management
- Inconsistency across locations: Diverse hardware, network, and compliance requirements lead to operational disparities and deployment inconsistencies.
- Manual lifecycle bottlenecks: Labor-intensive processes for hardware, OS, Kubernetes, and application operations become unsustainable at scale.
- Fragmented visibility: Lack of centralized data aggregation hinders enterprise-wide insights into firmware status, capacity, and issue identification.
- Scalability limitations: Orchestration tools designed for single data centers fail to scale efficiently for thousands of machines across hundreds of sites, leading to unreliable state synchronization and logistical challenges.
Core Architecture and Technologies
The solution employs a three-layer architecture: a centralized orchestration engine on AWS, distributed on-premises infrastructure (EKS Anywhere clusters), and hybrid connectivity (Direct Connect, Site-to-Site VPN). The AWS orchestration engine is built on serverless technologies to provide scalability and reduce operational overhead.
- AWS Serverless Stack: Utilizes AWS Lambda for processing requests, AWS Step Functions for workflow orchestration (with retry logic, error handling, and state checkpointing), Amazon DynamoDB as a centralized state repository, and Amazon EventBridge for event routing. AWS CodeBuild, AWS Batch, and AWS Systems Manager extend capabilities for specific tasks.
- Redfish APIs: Standard protocol for vendor-agnostic hardware management, enabling uniform control over BIOS, firmware, power, and health monitoring of bare-metal servers.
- Amazon EKS Anywhere: Deploys and operates Kubernetes clusters on on-premises hardware, leveraging the EKS Distro for consistency with cloud EKS deployments.
Inventory Management System
Single Source of Truth
Amazon DynamoDB acts as the central state repository for the Inventory Management System. It tracks sites, servers (with BIOS/firmware, IP, MAC, physical location), EKS Anywhere clusters (Kubernetes configs, node groups, add-ons), and operational orders. This system ensures that all infrastructure changes are reflected, providing a single, consistent source of truth for resource relationships and operational history across the hybrid environment.
Event-Driven Orchestration Engine
The orchestration engine is API-driven and event-driven. An Amazon API Gateway layer exposes RESTful APIs for CRUD operations, integrated with a unified operator portal. Incoming requests trigger Lambda functions, which validate parameters and interact with the order management system. Orders initiate workflows managed by AWS Step Functions, which then coordinate operations across on-premises infrastructure via hybrid connectivity. This design pattern ensures robust, scalable, and auditable infrastructure lifecycle management.