Case Study: Top Ten US Bank achieves real-time fraud detection and 5,000 TPS throughput with Hazelcast

Hazelcast IMDG Powers Real-Time Fraud Detection

Top Ten US Bank was hitting transaction-rate limits on its legacy relational database when running real-time fraud-detection algorithms, causing SLA breaches and blocking new business. To address this, the bank partnered with Hazelcast and adopted Hazelcast IMDG (with Hazelcast Jet for in-memory stream and batch processing) to support low-latency, always-on fraud checks.

Hazelcast moved 2TB (planned to grow to 5TB) of customer data into memory, ran fraud rules in situ, and added WAN replication for high availability; the deployment processes 5,000 transactions per second today with expectations to reach 10,000 TPS, meets tight SLAs, and preserves service during regional outages. By using Hazelcast, the bank eliminated the relational database bottleneck and achieved much lower latency for scoring and detection.