© 2018 Strange Loop
Real-time processing of transactions at MasterCard is not a nice to have, it is a must. And with credit card fraud running rampant, we have to be smarter and faster than the fraudster to protect our cardholders. In real-time, MasterCard needs to identify fraud real-time and decline a transaction before any exposure takes place. Oh, and by the way, this work is measured in milliseconds and occurs tens of thousands of times per second. To accomplish this is no small feat. From application, network, disk and memory i/o, java garbage collections and a myriad other issues, we have found solutions to handle this volume quickly with a high rate of success. I will discuss lessons learned and techniques regarding architecting systems for performance, how to create and implement a performance testing strategy. This strategy includes isolating tiers and testing to find their breaking point, monitoring systems to find issues, creating a capacity and demand model for the multiple tiers to find out their effect on each type of hardware and how it might cause issues with other tiers, understand Java garbage collection strategies, and more.
I have been a Senior Technical Consultant at MasterCard Worldwide for nine years. I have worked on application architecture for a majority of my career with a recent focus on high-volume, low latency performance engineering.