Enterprise Integration
Connecting systems that were never designed to talk to each other, with failure handled as a normal case.
Enquire about this serviceIntegration work where the interesting problems are the failure modes: partial writes, duplicate deliveries, upstream timeouts and reconciliation after an outage.
The design assumes the other system will be unavailable, will return something unexpected, and will occasionally process the same message twice. Idempotency, retry semantics and reconciliation are part of the design rather than the incident response.
Engagement process
- Mapping — the systems, their contracts and their real behaviour
- Failure analysis — what breaks, how often, and what it costs
- Design — idempotency, retry semantics and reconciliation
- Implementation with observability built in
- Verification against realistic failure scenarios