Equally True: Importance of testing with large data sets - Blog about tech, design, programming and mobile: "Surprisingly after running the tests with millions of data points something started breaking, and the CPU went up. This was the point I started realizing the true value of those algorithm classes that seemed theoretical at the time. Additionally, I realized its really hard to catch all race conditions in multithreaded applications with unit tests. Later on I did manage to catch the concurrency bugs after some debugging and improve the performance to acceptable level, but it opened my eyes for the importance of testing and training with realistic and large data sets.
'via Blog this'