Deployment Stamp Pattern
The Deployment Stamp pattern is a cloud architecture strategy for deploying multiple independent copies of an application and its entire supporting infrastructure. Each self-contained copy is called a stamp (or sometimes a scale unit, service unit, or cell). This pattern is primarily used to improve scalability, provide strong tenant isolation, and facilitate deployments across multiple geographic regions.
Instead of scaling up, the solution is scaled out by adding more stamps as demand grows. A global routing layer directs traffic to the appropriate stamp, but otherwise, each stamp is autonomous.
How It Works
The core principle is to define a complete, self-contained application stack and then replicate it.
- Define the Stamp: A single stamp’s architecture is defined using Infrastructure as Code (IaC). This includes all necessary resources: web servers, application services, databases, message queues, etc.
- Provision Stamps: Multiple instances of this stamp are provisioned. Each stamp is identical in its composition but is configured for its specific context (e.g., which tenants it serves).
- Route Traffic: A global routing component sits in front of all the stamps. It inspects incoming requests and uses metadata (like a tenant ID from a JWT, a subdomain, or a geographic header) to direct the user to their assigned stamp. This router is often a combination of DNS-based routing and a Layer 7 load balancer.
- Manage and Monitor: A centralized management plane is used to orchestrate deployments, updates, and monitoring across all stamps, ensuring consistency and operational visibility.
The Critical Role of Infrastructure as Code
The Deployment Stamp pattern is not merely enhanced by Infrastructure as Code (IaC); it is fundamentally enabled by it. Attempting to manage multiple stamps manually is impractical and defeats the purpose of the pattern, which relies on consistency and automation.
IaC serves as the master blueprint for a stamp. This blueprint, written in a declarative language like Terraform, Bicep, or CloudFormation, provides several key guarantees:
- Repeatability and Consistency: IaC ensures that every stamp provisioned from the blueprint is identical. This eliminates “configuration drift,” where manual changes cause environments to diverge over time, leading to unpredictable behavior.
- Automated Provisioning: When the system needs to scale, a new stamp can be provisioned automatically by executing the IaC template. This allows for rapid, on-demand scaling without manual intervention, reducing both lead time and the risk of human error.
- Centralized and Version-Controlled Updates: When a change is required—such as updating a service, patching a vulnerability, or modifying a resource configuration—the change is made once in the IaC source code. This change can then be reviewed, tested, and rolled out across all stamps in a controlled manner (e.g., using a ring-based deployment). This provides a single source of truth for the entire architecture.
- Lifecycle Management: IaC tools manage the full lifecycle of the infrastructure, including not only creation but also updates and, critically, decommissioning. This ensures that resources are cleaned up properly when a stamp is no longer needed.
graph TD subgraph "Global Infrastructure" direction LR Router(Global Router / Traffic Manager) end subgraph "Stamp 1 (e.g., EU Region, Tenants 1-100)" direction TB S1_WebApp(Web App) S1_API(API Service) S1_DB(Database) S1_Queue(Message Queue) S1_WebApp --> S1_API S1_API --> S1_DB S1_API --> S1_Queue end subgraph "Stamp 2 (e.g., US Region, Tenants 101-200)" direction TB S2_WebApp(Web App) S2_API(API Service) S2_DB(Database) S2_Queue(Message Queue) S2_WebApp --> S2_API S2_API --> S2_DB S2_API --> S2_Queue end subgraph "Stamp N (New Tenants)" direction TB SN_WebApp(Web App) SN_API(API Service) SN_DB(Database) SN_Queue(Message Queue) SN_WebApp --> SN_API SN_API --> SN_DB SN_API --> SN_Queue end User(User Request) --> Router Router -- "Tenant 10" --> S1_WebApp Router -- "Tenant 150" --> S2_WebApp Router -- "Tenant 210" --> SN_WebApp style S1_WebApp fill:#cde4ff style S2_WebApp fill:#cde4ff style SN_WebApp fill:#cde4ff
Description: A global router directs user traffic to the appropriate independent stamp based on tenant or regional information. Each stamp contains a full, isolated copy of the application stack.
Use Cases
- High Scalability for SaaS: When a multi-tenant application needs to scale to a very large number of customers with predictable performance and cost per tenant.
- Strong Tenant Isolation: For applications requiring strict data or performance isolation between tenants, such as in healthcare (HIPAA) or finance (PCI compliance). An issue in one stamp (e.g., a noisy neighbor) does not affect tenants in other stamps.
- Geographic Distribution (Geo-fencing): To deploy instances in specific regions to reduce latency for users in that region or to comply with data residency laws (e.g., GDPR). For extreme availability and active-active global distribution, this approach can be evolved into the Geode pattern.
- Controlled, Ring-Based Rollouts: New versions of the application can be deployed to a single stamp (a “ring”) serving a small group of early adopters or internal users before being rolled out to all stamps, minimizing the blast radius of a bad deployment.
- Disaster Recovery: Stamps can be deployed across different availability zones or regions, providing a high degree of fault tolerance.
Implementation on Cloud Platforms
While the pattern is platform-agnostic, its implementation varies slightly across major cloud providers:
-
Microsoft Azure: The pattern is a first-class concept in many Azure services.
- Stamp Definition: Azure Resource Manager (ARM) templates or Bicep files are used to define the stamp’s resources.
- Routing: Azure Front Door or Azure Traffic Manager can act as the global router.
- Orchestration: Azure DevOps Pipelines or GitHub Actions are used to automate the provisioning and updating of stamps.
-
Amazon Web Services (AWS):
- Stamp Definition: AWS CloudFormation or Terraform is used to define a stamp’s resources within a VPC (e.g., EC2 instances in an Auto Scaling Group, an RDS database, SQS queues).
- Routing: Amazon Route 53 can be used for DNS-based routing, or an Application Load Balancer can inspect requests and route them to the correct target group representing a stamp.
- Orchestration: AWS Control Tower or custom scripts with AWS Systems Manager can be used to manage deployments across multiple AWS accounts or regions, where each account/region can represent a stamp.
-
Google Cloud Platform (GCP):
- Stamp Definition: Google Cloud Deployment Manager or Terraform defines the stamp’s resources (e.g., GKE clusters, Cloud SQL instances, Pub/Sub topics).
- Routing: Cloud Load Balancing can serve as the global router, directing traffic based on host, path, or headers.
- Orchestration: Cloud Build can be used to automate the deployment and management of stamps across different projects or regions.
Challenges and Considerations
- Deployment Complexity: This pattern is entirely dependent on mature and robust Infrastructure as Code (IaC) practices. Managing and updating a large number of stamps requires a very high degree of automation, as manual management is not feasible.
- Cost: The pattern can be more expensive at a small scale because it duplicates infrastructure that might otherwise be shared. However, it offers more predictable, linear cost scaling as the user base grows.
- Cross-Stamp Operations: Performing operations that span all tenants, such as global analytics or reporting, is complex. This often requires a separate system to aggregate data from all stamps into a centralized data warehouse.
- Tenant Migration: Moving a tenant from one stamp to another (e.g., for rebalancing) is a complex operation that requires careful planning and custom tooling to migrate data and update routing rules without downtime.
- Global Component Management: The global router is a critical component and a potential single point of failure. It must be designed for high availability and resilience.
Resources & links
Articles
-
Deployment Stamp pattern - Microsoft Azure
The official documentation from the Azure Architecture Center, which provides a comprehensive overview of the pattern, its use cases, and key considerations. It is the foundational reference for this pattern.
-
Deployment Stamps 101 (Medium)
A practical article that explains the concept with clear diagrams and discusses the “why” behind the pattern, focusing on scalability, resilience, and cost-effectiveness.