The Service-Based Architecture Paradigm
When To Use
- Multi-team organizations with domain-aligned services
- Systems requiring independent deployment of components
When NOT To Use
- Single-team projects small enough for a monolith
- Latency-sensitive systems where inter-service calls are prohibitive
When to Employ This Paradigm
- When teams require a degree of deployment independence but are not yet prepared for the complexity of managing numerous microservices.
- When shared databases or large-scale systems (like ERPs) make full service autonomy unrealistic.
- When establishing clear service contracts for partner teams or external consumers.
Adoption Steps
- Group Capabilities: Bundle related business functions into a small set of well-defined services, each with a designated owner.
- Define Service Contracts: Publish formal specifications using standards like OpenAPI or AsyncAPI, including Service Level Agreements (SLAs) and a clear versioning strategy.
- Control Database Schemas: Even when services share a database, assign explicit ownership for each schema or table. Gate all breaking changes through a formal review process.
- Establish Service Mediation: Use a service registry or an API gateway to handle routing, authentication, and observability.
- Plan for Evolution: Identify architectural "hotspots" that are likely candidates for being split into more granular services in the future.
Key Deliverables
- An Architecture Decision Record (ADR) that outlines service boundaries, data ownership rules, and coordination mechanisms.
- A suite of contract tests and consumer-driven contract tests for each service to validate stability.
- Runbooks that describe deployment procedures, rollback plans, and service dependencies.
Risks & Mitigations
- Coupling Through a Shared Database:
- Mitigation: Changes to a shared database can have cascading effects across services. Mitigate this by using database views, replication, or a formal schema deprecation schedule to manage change.
- Architectural Degradation:
- Mitigation: Without strong governance, this architecture can degrade into a "distributed monolith": a monolith with the added complexity of network hops. Track coupling metrics closely and enforce strict ownership of services and data to prevent this.
Concrete Components
Vocabulary for the tools and abstractions an implementation of this
paradigm tends to carry. Not dependencies, and not tools: frontmatter.
api-gateway: single ingress that routes to coarse-grained services and centralizes cross-cutting concernsservice-registry: directory of available services with health status and contractsschema-management: shared schema repo for types crossing service boundaries
Exit Criteria
- An ADR outlines service boundaries, explicit data ownership per schema or table, and the coordination mechanism (service registry or API gateway) before any service is deployed independently.
- Formal service contracts (OpenAPI or AsyncAPI) with versioning and SLA statements exist for every inter-service interaction point.
- Consumer-driven contract tests exist for each service and pass in CI before a service owner merges breaking interface changes.
- Shared-database coupling hotspots are documented with a schema deprecation schedule; any service reading another service's tables without a view or replication layer is flagged as an architectural violation.
