OpenShift & Kubernetes for financial systems

By Cédric Barme · Founder of NSY

Containerising a critical banking or insurance application has little in common with deploying a stateless microservice. Sessions, XA transactions, messaging, auditability requirements: NSY designs and hardens OpenShift and Kubernetes platforms that respect these constraints — instead of discovering them in production.

What "containerising legacy" really means

The standard narrative assumes twelve-factor, stateless applications that restart at will. The real estate of financial institutions is made of Java EE applications with replicated sessions, persistent JMS queues, scheduled batches and startup dependencies. Containerising them demands specific architecture work: liveness and readiness probes that reflect real state, zero-downtime upgrade strategies, volume and secret management, dependency ordering. That bridge between the two worlds is exactly what NSY builds — see also Java EE platform migration.

Typical engagements

• Containerisation trajectory — which applications, in which order, with what return: not all of them deserve the journey, and saying so is part of the job.
• Platform architecture — namespaces and segregation, networking (ingress, service mesh where justified), persistent storage, image registries and software supply chain.
• Operations and reliability — observability (metrics, logs, traces), capacity management, incident post-mortems and durable remediation plans.
• Private and hybrid cloud — OpenShift on internal infrastructure, controlled extension to AWS/Azure/GCP where data residency and reversibility allow.

Compliance built into the architecture

In a regulated environment the platform must demonstrate: who deployed what, when, with which image, approved by whom. Change traceability, role separation, data residency and provider reversibility — ACPR/AMF requirements that DORA reinforces further. Bolting these on afterwards costs a multiple of designing them in.

Why NSY

A software engineer since 2012, Cédric Barme has spent his career on both sides: the critical applications that run on these platforms, and the platforms that carry them. That dual perspective avoids the classic failure — a platform that meets every standard but that real applications cannot use. Single point of contact, no pyramid, three clients maximum. The FAQ details the engagement model.