Service Discovery 101: The Phonebook for Distributed Systems

Consider how a food delivery application works when someone places an order.

The app needs to know which restaurants are currently open, where they’re located, whether they’re accepting orders, and how busy they are. If a restaurant closes early or temporarily stops accepting orders due to high demand, the app needs to know immediately. This information changes constantly throughout the day, and having outdated information leads to failed orders and frustrated customers.

Modern software applications face remarkably similar challenges.

Instead of restaurants, we have services. Instead of checking if a restaurant is open, we need to know if a payment service is healthy and responding. Instead of finding the nearest restaurant location, we need to discover which server instance can handle our request with the lowest latency. Just as the food delivery app would fail if it sent orders to closed restaurants, our applications fail when they try to communicate with services at outdated addresses or send requests to unhealthy instances.

In today’s cloud environments, applications are commonly broken down into dozens or even hundreds of services. These services run across a dynamic infrastructure where servers appear and disappear based on traffic, containers restart in different locations, and entire service instances can move between machines for load balancing.

The fundamental question becomes: how do all these services find and communicate with each other in this constantly shifting landscape?

In this article, we look at service discovery, the critical system that answers this question. We’ll examine why traditional approaches break down at scale, understand the core concepts, and patterns of service discovery.

Life Before Service Discovery