Quantifying scalability with the Universal Scalability Law

Distributed systems used to be the exception, but today they're the norm, so it's more useful than ever to be able to quantify scalability. Baron Schwartz explains how to use the Universal Scalability Law to characterize how your systems truly behave, why they don't scale like they should, and how to improve them. It's a simple, elegant solution, and, although formal, it requires no math.

You’ve probably heard a lot of people talk about “linearly scalable” systems, but what does that mean? It turns out it’s an equation of throughput under load: the Universal Scalability Law. All engineers, architects, and operators should be familiar with this simple and fundamental concept. When you understand how to model systems by measuring their behavior, you can do amazing things, such as determine how close to linear they are; understand the two causes of sublinearity and measure which dominates; understand what types of things you can improve about them; forecast beyond what you can measure (capacity planning); and avoid building bottlenecks in your systems at all. Baron Schwartz explains how to use the Universal Scalability Law to characterize how your systems truly behave, why they don’t scale like they should, and how to improve them. You’ll see why systems degrade under load and learn how to apply the lessons the USL teaches—all without even doing math (although you can if you want to). This intuitive concept will forever change the way you think about systems. And as a bonus, once you see it, you’ll never be able to unsee it, so you’ll find yourself using it constantly in all kinds of situations.