Example: How to scale round robin streaming applications

Use Case

A User receives messages through a website in real-time (Push instead of Pull). Examples: Chats, Stock Exchange, Notifications.

Given

• A user browsing a website
• A non-user aware gateway that equally distributes user requests to a backend API (round robin).
• A backend API that is subscribed to event topics on message broker can be scaled out dynamically.
• A non-user aware message broker that equally publishes messages to topics (round robin).

Challenge

In this publish subscribe scenario the challenge is to router the messages for the user to those backend nodes holding an open connection.

Solution Approach: Shared state via Hazelcast

Prerequisites

• Docker
• Docker Compose
• Gradle

Build

cd websocket-backend
./gradlew build bootJar && docker build --tag websocket-backend .
cd ..

Run

docker-compose up