Case Study: Mastercard achieves sub-second fraud detection at billion+ transactions per day with VMware Tanzu

How Mastercard fights fraud with Apache Geode

Mastercard’s Decision Management Platform (DMP) validates every card transaction for nearly 200 million cardholders, so preventing fraud at scale is mission-critical. Faced with growing card usage, terabytes of time-aware data (over 20 billion aggregates), and strict SLAs—50 ms average latency at 60,000 TPS—Mastercard needed a way to process billions of transactions without sacrificing performance or consistency.

They adopted Apache Geode’s in-memory architecture and applied patterns like data co-location, balanced partitioning, byte-array storage, server-side functions, delta updates and extensive aggregation. The result: a 40 TB Geode cluster delivering sub-millisecond reads and ~8 million reads/sec, meeting latency targets at massive throughput; context-aware functions cut client-server traffic by 95%, lowered latency by ~50% and reduced server CPU by ~40%, while delta propagation halved peer-to-peer network traffic.

MasterCard

Andrey Zolotov

Software Engineer