Performance Engineering At MasterCard
summary
Claim: MasterCard builds a Safety Net real-time fraud detection pipeline that handles tens of thousands of transactions per second with responses in tens of milliseconds. Engineering specifics: isolate and stress individual tiers (messaging, rules engine, NoSQL) to find per-tier limits, discovered an MQ bottleneck at ~80k msgs/s per queue, measured 10,000 full GC pauses/day and six hours lost to GC, and eliminate long pauses by adopting Zing JVM. Capacity planning uses a heat‑map capacity model and Splunk-backed dashboards.
AI-written summary · about
chapters (16)
- 0:00Intro
- 3:48Three Guiding Principles
- 5:00Tearing
- 8:08Test Environment
- 10:55Testing Tiers
- 12:13Stress Testing
- 14:31Tracking Metrics
- 19:39Testing Results
- 21:27Code Issues
- 22:35Garbage Collection
- 25:10Zing
- 26:10Strategic
- 29:39Heatmaps
- 31:34Dashboards
- 32:36Steps for Success
- 34:47Questions Comments
related talks
24mBuilding Planet-Scale Payment Platforms
51mReal-time Fraud Detection: Challenges and Solutions
13mBuilding Financial Fortresses: Real-Time Fraud Detection
15mPlatform Engineering for FinTech
09mPerformance is the first UX challenge to consider
39mThe Science Behind Delightful User Experience
40mPerformance Testing and Big Data
29mGuardians of the Gateway: Battling SMS Fraud in Real Time