Quarkus hosting on a small, fast server

Why Quarkus suits a VPS

Quarkus moves work from startup to build time. The result is faster startup and a smaller memory footprint than a typical Spring or Jakarta EE application. On a VPS that means you can run more services in the same RAM, or pick a smaller plan for the same workload.

Two ways to run it

JVM mode (fast-jar)

The default mvn package produces target/quarkus-app/. You start it with java -jar quarkus-run.jar. This is the simplest option and it works with every extension. We run it as a systemd service, with the settings in an environment file, for example QUARKUS_HTTP_PORT and QUARKUS_HTTP_HOST=127.0.0.1.

Native executable

A native build uses GraalVM and starts in milliseconds with very little memory. Building it needs a lot of RAM, so build in CI and upload the executable. Native executables are Linux x86-64 binaries, so build for that target. Some libraries need extra configuration for native mode.

What we set up

• OpenJDK 17 or 21 for the JVM build, or a plain runtime directory for a native binary.
• Your application as a systemd service under its own user, restarting on failure.
• Nginx with free SSL in front, and Quarkus listening on localhost.
• Optional PostgreSQL or MySQL on the same server.
• The deploy-app command for new versions.

Memory guidance

A Quarkus REST service in JVM mode runs well with a heap of 128 to 256 MB. Native mode often stays under 100 MB of resident memory. That is why the 2 GB plan can host a Quarkus service, a PostgreSQL database and Nginx with room to spare.