jarz-streaming | Efficient Java Deployment

The "Fat JAR" Bottleneck

Traditional Java deployments ship monolithic binaries. If you change 200KB of code in a 100MB application, you still ship 100MB. This results in massive storage redundancy and network saturation. jarz-streaming changes the paradigm to object-based streaming.

Storage & Transfer Impact

*Data assumes a typical Spring Boot App (100MB) with frequent small logic updates.

📦

Storage Redundancy

99.8% of a typical update is duplicate data (libraries like `spring-core`, `log4j`). We deduplicate this by bundling dependencies in optimized blocks in S3-compatible storage.

📡

Network Saturation

Deploying "Fat JARs" to edge nodes or air-gapped environments chokes bandwidth. Streaming sends only the ~50KB Index + specific required class blocks.

⚡

The Solution: Block Streaming

Treat class bundles as content-addressable objects. The JVM boots from a lightweight Index and lazy-loads only the execution path.

Architecture Explorer

Click through the phases to understand how jarz-streaming transforms the Build, Registry, and Runtime lifecycles.

1. Build Phase

jarz-maven-plugins

Bundle, Hash, Cross-Reference & Upload.

2. Registry Phase

jarz-registry

Central hub for Artifacts & Metadata.

3. Runtime Phase

jarz-streaming

Lazy ClassLoader & Block Streaming.

🔨

Build Phase Analysis

During the build, the plugin analyzes the application classpath. Instead of zipping classes into a monolithic JAR, it:

Analyzes cross-references between classes.
Creates optimized bundles containing classes and their immediate dependencies.
Uploads unique bundles to the central storage backend (maven/gradle or S3 compatible).

Optimization: Dependencies are downloaded in efficient blocks rather than individually, reducing HTTP overhead.

Air-Gapped & Disconnected Environments

Designed specifically for Defense, Telco, and Banking sectors where public internet access is restricted. jarz-streaming utilizes an "Inner CDN" model to ensure secure, low-latency distribution.

🔒 Secure Origin

Validated with S3-compatible object storage acting as the primary origin within the secure perimeter.

🏎️ Inner CDN

Uses standard HTTP/2 proxies (Nginx/Envoy) as edge caches. Streams are read-only and hash-verified against the Index JAR.

🔄 Race-Free Invalidation

Content-Addressable Storage (CAS) ensures no overwrites. Old objects are orphaned; new versions get new hashes. Zero cache collisions.

Secure Perimeter Visualization

🛢️

S3 Storage (Origin)

Object Bundles & Hashes

🌐

Inner CDN (Proxy)

HTTP/2 Edge Cache

☕

JVM (Lazy Loader)

Requests Needed Bundles

Sample Nginx Caching Proxy

# Inner CDN Configuration

proxy_cache_path /var/cache/jarz 
    levels=1:2 
    keys_zone=jarz_zone:10m 
    max_size=10g 
    inactive=1y;

server {
    listen 443 ssl http2;
    server_name jarz-proxy.internal;

    location / {
        # Upstream: S3 Compatible Storage
        proxy_pass https://s3-backend;
        
        # Caching Rules for Immutable Blocks
        proxy_cache jarz_zone;
        # Cache hits for 1 year (Immutable)
        proxy_cache_valid 200 1y; 
        
        # Optimization
        proxy_buffering on;
        proxy_ignore_headers Expires Cache-Control;
    }
}

ROI Calculator

Estimate your savings based on fleet size and deployment frequency.

Artifact Size (MB)

50MB 100 MB 500MB

Nodes / Instances

10 100 Nodes 1000

Deploys per Week

1 10 / week 50

Projected Bandwidth Savings

0 TB

per year

Cumulative Bandwidth Usage (Year 1)

Technical Specifications

Java Version	JDK 11 or higher (Required for custom ClassLoader API)
Build Tool	Maven 3.8+ (Gradle plugin in beta)
Storage Protocol	S3 API (Validated on AWS S3, MinIO, etc.)
Transport	HTTP/1.1 or HTTP/2 (Recommended for multiplexing)

Ready to shape the cloud?

Contact Ecosystem Team →