Everything I've covered so far about Relay is stuff that's in the browser. Relay needs to send it's GraphQL queries somewhere. For this, you need a GraphQL backend. You can implement this using Node.js and a handful of GraphQL libraries. You create what's called a schema, describing all the datatypes, queries, and mutations that will be used.

In the browser, Relay helps you scale your applications by reducing data-flow complexity. You have a means to declare what data is needed, without worrying about how it is fetched. It's the schema in the backend that actually needs to resolve this data.

This is another scaling problem that GraphQL helps address. Modern web applications are composed out of microservices. These are smaller, self-contained API endpoints that serve some particular purpose that's smaller than an entire app (hence the term micro). It's the job of our application to stitch together these microservices and provide the frontend with meaningful data.

Again, you're faced with a scalability issue—how do you maintain a backend that's composed out of many microservices without introducing insurmountable complexity? This is something that GraphQL types excel at. In the following chapter, you'll begin the implementation of your Todo application with the backend GraphQL service.