Building Microservices, 2nd Edition

Chapter 13. Scaling

“You’re gonna need a bigger boat.”

Chief Brody, Jaws

When we scale our systems, we do so for one of two reasons. Firstly, it allows us to improve the performance of our system, perhaps by allowing us to handle more load or by improving latency. Secondly, we can scale our system to improve its robustness. In this chapter, we’ll look at a model to describe the different types of scaling, and then we’ll look in detail at how each type of scaling can be implemented using a microservice architecture. At the end of this chapter, you should have an array of techniques to handle the scaling issues that may come your way.

To start with, though, let’s look at the different types of scaling you might want to apply.

The Four Axes of Scaling

There isn’t one right way to scale a system, as the technique used will depend on the type of constraint you might have. We have a number of different types of scaling we can bring to bear to help with performance, robustness, or perhaps both. A model I’ve often used to describe the different types of scaling is the Scale Cube from The Art of Scalability,¹ which breaks scaling down into three categories that in the context of computer systems cover functional decomposition, horizontal duplication, and data partitioning. The value of this model is that it helps you understand that you can scale a system along one, two, or all three of these axes, depending on your needs. However, especially in a world of virtualized infrastructure, ...