Comparing jlink Images to Traditional Deployments
What you'll learn: How custom runtime images created with jlink stack up against traditional full JDK distributions in terms of size, performance, and trade-offs.
When you distribute a Java application, you traditionally bundle it with a full JRE or JDK, which weighs hundreds of megabytes. With jlink, you create a custom runtime image containing only the modules your application actually uses—like packing for a trip with just the essentials instead of your entire wardrobe.
Size Comparisons
A full JDK distribution typically ranges from 150-300+ MB depending on the platform. A jlink image for a modest modular application might be 30-50 MB, and minimal "Hello World" images can shrink below 30 MB with compression. This dramatic reduction matters for:
- Container deployments (smaller Docker images)
- Download times for client applications
- Storage costs in cloud environments
Startup Time Benefits
Custom runtime images offer faster startup because:
- The JVM loads fewer platform modules
- Class Data Sharing (CDS) can be pre-configured
- Reduced classpath scanning overhead
Real-world improvements range from 10-30% faster cold starts, significant for serverless functions and microservices.
Trade-offs
Flexibility vs Optimization: A full JDK runs any Java application; a jlink image runs only the modules you included. Adding functionality later requires rebuilding the image.
Maintenance burden: You manage separate runtime images per application, rather than a single shared JDK installation.
Debugging complexity: If you exclude jdk.compiler or debugging tools from your image, troubleshooting production issues becomes harder.
Platform specificity: jlink images are platform-dependent; distributing to Windows, Linux, and macOS requires three separate builds.
Key Takeaway:
jlink images drastically reduce size and improve startup times compared to full JDK distributions, but trade universal flexibility for application-specific optimization—choose them when deployment efficiency outweighs runtime flexibility needs.