Beyond Tomcat: Which Spring Boot Embedded Server—Jetty or Undertow—Delivers the Best Performance?

What is an Embedded Server?

In a Spring Boot application the web server is packaged inside the executable JAR and started together with the application, eliminating the need for a separate servlet container installation.

Embedded Server = Web server packaged inside the application and started together with it.

The JAR contains the application code and one of the three supported servlet containers: Tomcat, Jetty, or Undertow.

Why Use Embedded Servers?

Embedding the server removes complex environment dependencies, reduces deployment steps, simplifies scaling, and lowers operational cost, especially in micro‑service architectures. It enables a single‑command start, no extra server installation, and seamless Docker/Kubernetes integration.

Spring Boot’s Three Embedded Servers

Tomcat – stable, mature; suited for general‑purpose business systems.

Jetty – lightweight, flexible; suited for long‑lived connections and streaming.

Undertow – non‑blocking, high‑performance; suited for high‑concurrency, low‑latency APIs.

Default Tomcat Configuration

Adding the spring-boot-starter-web dependency pulls in Tomcat, auto‑configures the servlet container, and starts an HTTP service.

<dependency>
    <groupId>org.springframework.boot</groupId>
    <artifactId>spring-boot-starter-web</artifactId>
</dependency>

Switching to Jetty

Jetty offers lower memory usage, better WebSocket/long‑connection support, and faster startup.

Exclude Tomcat.

Add the Jetty starter.

<dependencies>
    <!-- Exclude Tomcat -->
    <dependency>
        <groupId>org.springframework.boot</groupId>
        <artifactId>spring-boot-starter-web</artifactId>
        <exclusions>
            <exclusion>
                <groupId>org.springframework.boot</groupId>
                <artifactId>spring-boot-starter-tomcat</artifactId>
            </exclusion>
        </exclusions>
    </dependency>
    <!-- Add Jetty -->
    <dependency>
        <groupId>org.springframework.boot</groupId>
        <artifactId>spring-boot-starter-jetty</artifactId>
    </dependency>
</dependencies>

Switching to Undertow

Undertow uses a non‑blocking I/O model and an event‑driven architecture, allowing a few threads to handle many requests.

<dependencies>
    <!-- Exclude Tomcat -->
    <dependency>
        <groupId>org.springframework.boot</groupId>
        <artifactId>spring-boot-starter-web</artifactId>
        <exclusions>
            <exclusion>
                <groupId>org.springframework.boot</groupId>
                <artifactId>spring-boot-starter-tomcat</artifactId>
            </exclusion>
        </exclusions>
    </dependency>
    <!-- Add Undertow -->
    <dependency>
        <groupId>org.springframework.boot</groupId>
        <artifactId>spring-boot-starter-undertow</artifactId>
    </dependency>
</dependencies>

Common Configuration Parameters

Configuration can be done via application.properties or application.yml regardless of the chosen server.

server.port=8081
server.servlet.context-path=/api
server.tomcat.max-threads=200

server:
  port: 8081
  servlet:
    context-path: /api
  tomcat:
    max-threads: 200

For Jetty or Undertow replace the tomcat prefix with jetty or undertow respectively.

Project Structure Example

/opt/projects/embedded-server-demo
├── pom.xml
├── src
│   ├── main
│   │   ├── java/com/icoderoad/Application.java
│   │   └── resources/application.yml
│   └── test

Server Selection Guidance

General business systems: Tomcat (stable, default).

High‑concurrency API gateways: Undertow (best performance).

Long‑connection / real‑time communication: Jetty (strong connection management).

Performance Core Comparison

Tomcat allocates one thread per request, while Undertow can handle many requests with a small number of threads, reducing thread‑context overhead in high‑concurrency scenarios.

Conclusion

Performance bottlenecks often stem from the default embedded server rather than application code. Tomcat is reliable, but Jetty and Undertow provide distinct advantages for specific workloads. Engineers should evaluate the default choice and switch when the workload demands better scalability or lower latency.