Decoupling Data Domains at Auxin with Event Sourcing

How we used event modeling and event sourcing to create a loosely coupled, highly scalable data architecture for the Auxin data hub.

The challenge: sharing data without coupling services

As our portfolio of apps grew (each an independent service with its own DB and webapp), sharing data across domains became essential. I pushed back on service‑to‑service calls because they create tight coupling and cascading failures. We needed a way for services to learn from the same facts without sharing mutable state.

The solution: thinking in events

Instead of modeling around state, we modeled around immutable facts: events.

  • SensorReadingReceived
  • DeviceConfigurationUpdated
  • NutrientRecipeCreated

We adopted EventStoreDB (now Kurrent) and treated the append‑only event log as the fundamental data structure.

How it works at Auxin

  1. Append‑only streams are the source of truth. Raw domain events are written immutably. The stream history is the canonical record.
  2. Materialized views per service. Most services project from the event store into their own read models optimized for their use cases.
  3. Duplicate storage by design. Multiple services may keep similar views. Storage is cheap. Correctness comes from a single primary data source, the event store.
  4. No direct calls between services. Services subscribe to the events they care about instead of calling each other. This preserves autonomy and failure isolation.

Working with an event store has a learning curve: event modeling, projections, idempotency. The payoff is clarity, auditability, and evolvability.

Business impact at Auxin

  • Loose coupling at the data layer. Teams move faster without coordinating releases.
  • A reliable hub for the sector. Third‑party apps can plug in by producing or consuming events without getting entangled in our internals.
  • End‑to‑end traceability. The event log provides a full audit trail for compliance and root‑cause analysis.
  • Better analytics and AI. Unified events enable cross‑domain insights while each service stays independent.

Context and further reading