Case Study: BNP Paribas achieves in-memory speed to meet strict customer-facing SLAs with Hazelcast

How BNP Paribas Bank Polska was able to leverage in-memory speed to meet strict customer-facing SLAs

BNP Paribas (BGZ Group in Poland) needed to migrate and reconcile customer data from an acquired bank and modernize its Internet banking architecture to deliver millisecond response times without disrupting stable core systems. After evaluating in-memory technologies, the team chose Hazelcast (initially Hazelcast IMDG, later also Hazelcast Jet) to meet tight migration windows and aggressive customer-facing SLAs.

Hazelcast was deployed as an in-memory operational store and modern caching layer—used in a 16‑node POC for migration and in production clusters for Internet banking and streaming. The migration was completed over a weekend, reconciliation time dropped from ~4 hours (SQL) to 20 minutes, and production latency fell well below targets (required max 200 ms; Hazelcast delivered ~0.8 ms avg) while handling 84,000 requests/sec versus the 1,500/sec SLA. Real-time replication across nodes and data centers provided high availability and zero downtime, and BNP Paribas has since expanded Hazelcast across multiple clusters and use cases.

BNP Paribas

Szymon Domagala

Enterprise Architect