Kafka and Artemis JMS in financial environments
By Cédric Barme · Founder of NSY
In a financial system, messaging is not plumbing: it is what guarantees an order is neither lost nor executed twice. NSY designs and hardens these infrastructures — Kafka for streaming, Artemis JMS and MQ for transactional messaging — from trading platforms to insurance cores.
Two worlds, two tools
Kafka and JMS are not interchangeable, and many architectural drifts start with that confusion. Kafka excels at flows: high-throughput ingestion, replay, fan-out to multiple consumers, event sourcing. JMS messaging (Artemis, MQ) excels at transactions: guaranteed delivery to one consumer, XA transactions with the database, point-to-point queues. The right choice — and often the right combination — is decided on business invariants, not fashion.
Typical engagements
- Kafka architecture — partition sizing, delivery semantics (at-least-once, exactly-once and what it really costs), topic and schema governance, integration with core banking systems.
- Artemis / MQ high availability — clusters, replication, tested automatic failover; persistent queues that survive incidents instead of amplifying them.
- Legacy broker migration — progressive replacement of ageing brokers with Artemis or Kafka: parallel flow replication, dual consumption, consumer-by-consumer cutover — never a hard cut.
- Production diagnosis — growing queues, stalling consumers, poison messages: root-cause analysis and durable remediation, not ritual restarts.
Integration with the application platform
Messaging never lives alone: it integrates with application servers (WildFly / JBoss EAP, where Artemis is the embedded broker), containerised platforms (OpenShift / Kubernetes) and the traceability constraints of regulated environments. NSY works across the whole chain — which is what prevents siloed diagnoses where every layer blames its neighbour.
Why NSY
Cédric Barme built real-time trading and risk platforms in investment banking — systems where a lost message is measured in financial exposure. That standard has applied to every engagement since NSY was founded in 2018. The FAQ covers the use cases; for a specific context, reply within 48 business hours.
Describe your architecture — honest read within 48 business hours
Kafka, Artemis, MQ or all three: diagnosis first, commitment second.