Runtime coupling occurs when the availability of one component is impacted by the availability of another. For example, if a failure in one service causes a system-wide outage, it indicates high runtime coupling between the system’s components. Cascading runtime issues are often mitigated using resilience patterns.
Resilience Patterns
Resilience patterns help systems remain functional despite failures, high load, or unexpected conditions. These patterns are especially crucial in distributed systems and cloud architectures. Here are some of the most popular resilience patterns:
- Retry – Automatically retries a failed operation after a delay, often using exponential backoff to avoid overwhelming the struggling component.
- Circuit Breaker – Temporarily blocks requests to a failing component to prevent cascading failures and allows for gradual recovery.
- Bulkhead – Isolates failures in one part of the system to prevent them from affecting other components.
- Timeouts – Limits how long a request waits for a response, preventing hanging operations.
- Fail Fast – Immediately returns an error when a failure is detected instead of waiting for a timeout.
- Fallback – Provides a default response or alternative service when the primary service fails.
- Asynchronous Processing – Uses a message bus to decouple task production from task execution.
- Rate Limiting – Controls request rates to prevent system overload.
Runtime Coupling and Balanced Coupling
A tight runtime interdependency binds the lifecycles of affected components, making it difficult to evolve one without carefully considering its impact on the other. Lifecycle coupling is inversely proportional to the distance dimension. Therefore, reducing runtime coupling increases the distance between components. To keep coupling balanced, an increased distance must be compensated by reducing shared knowledge—lowering the components’ integration strength.