Introduction

Following the development of a scalable email-sending service using AWS SES, Spring Boot, and AWS Lambda, I set out to optimize its performance. The focus was to address the cold start latency and memory usage inherent to Java applications on AWS Lambda. To achieve this, I turned to GraalVM Native Image, a technology designed to compile Java applications into native executables. This article outlines the implementation and results of this optimization.

Why GraalVM Native Image?

GraalVM Native Image compiles Java applications ahead of time (AOT) into standalone executables. By doing so, it eliminates the need for a JVM at runtime, resulting in:

• Reduced Cold Start Times: Applications start almost instantly, a crucial factor for serverless environments.

• Lower Memory Usage: By stripping unnecessary components, it creates a lightweight application footprint.

These advantages make GraalVM an ideal solution for improving serverless application performance.

Implementation Steps

1. Project Setup

I began with AWS’s pet-store-native project, which provides a reference implementation for converting Spring Boot 3 applications into GraalVM-native images. This served as the foundation for integrating native image capabilities into the email-sending service.

2. Adapting for ARM Architecture

Since my environment uses an ARM-based architecture, the Dockerfile required modifications:

• Updated the base image to align with ARM.
• Configured the GraalVM compiler for ARM compatibility. These changes ensured the native image was optimized for the target system.

3. Runtime Configuration

Creating a custom bootstrap file for the runtime environment is essential to ensure the proper initialization and startup of the application. This file defines the entry point for the Lambda function and initializes the runtime environment. It also provides flexibility in configuring application parameters, enabling seamless integration with AWS Lambda.

I also enabled HTTP protocol support in the GraalVM Maven plugin and integrated the AWS Java Container for Spring Boot to handle API Gateway events. These configurations ensured the application could efficiently process HTTP requests and responses in its native image form.

4. Deploying the Application

Using the AWS Serverless Application Model (SAM), I deployed the native image as a Lambda function. Key customizations included:

• Switching from HTTP API Gateway to a standard API Gateway to enable API key-based authentication.
• Implementing usage plans for secure and scalable API access. These adjustments not only enhanced security but also allowed better resource allocation for the function.

Results

The transition to GraalVM Native Image yielded significant improvements:

• Cold Start Times: Reduced by eliminating JVM initialization.

• Memory Usage: Minimized due to the compact nature of native executables.

• Performance Scaling: Faster response times and better handling of concurrent requests.

Native image

SpringBoot3

Additionally, the API Gateway integration provided robust control over access and usage, enabling the service to function as a secure and scalable endpoint.

Lessons Learned

Through this implementation, I gained a deeper understanding of the interplay between GraalVM, Spring Boot, and AWS Lambda. The process highlighted the importance of:

• Optimizing for specific architectures to maximize performance.
• Configuring runtime environments to balance flexibility and efficiency.
• Leveraging tools like AWS SAM for streamlined deployment.

This project reinforced the potential of GraalVM Native Image as a powerful optimization tool for serverless Java applications, offering a compelling path forward for enhancing performance and reducing costs in production environments.

GitHub Repo for this project

Resources

Replatforming Java Applications with the Updated AWS Serverless Java Container

Quick Start Guide: Spring Boot 3

GraalVM Native Image: Faster, Smarter, Leaner

Going AOT: A Comprehensive Guide to GraalVM for Java Applications by Alina Yurenko | SpringIO

Going Native: Building Fast and Lightweight Spring Boot Applications with GraalVM by Alina Yurenko