🌟 The Asynchronous Engine (New in v1.6)

🏎️ Asynchronous Pipeline & Zero Head-of-Line Blocking

Forget lag spikes when loading chunks. ApexNetwork v1.6 completely removes Netty's single-thread bottleneck.

• Leap-Frogging: Massive chunk packets are automatically detoured off the main network thread into a parallel worker pool. Meanwhile, tiny critical packets (like sword swings, movement, and keep-alives) instantly bypass the chunk processing, arriving with zero delay.
• Crisp Interactions: You can be downloading gigabytes of chunk data in the background, and your item usage latency will remain perfectly at your baseline ping.

🗜️ Hybrid LZMA2 & Zstd Stack

Why choose one compression algorithm when you can have both?

• Smart Routing: The new HYBRID mode intelligently analyzes every single packet. Tiny packets are compressed instantly with the lightning-fast Zstd codec (10x faster than vanilla Zlib), while massive chunk packets are intercepted and crushed with maximum-efficiency LZMA2.
• Benchmark-Proven: Real-world testing on vanilla 1.21.1 shows Zstd lv9 + Dictionary achieving 7.36x compression at 2x the speed of Zlib, while Hybrid LZMA2 reaches 9.5x compression ratios (89.5% bandwidth saved) on large chunk payloads. Modded servers with higher data density can expect even greater savings.
• No Rubberbanding: Unlike other mods, you can safely crank compression to maximum settings. ApexNetwork's Asynchronous Pipeline offloads heavy compression to worker threads, so even LZMA2 lv9 or Zstd lv22 won't cause lag spikes or rubberbanding — your movement and combat packets always bypass the compression queue.

📖 Automatic Dictionary Synchronization (New in v1.6.1)

No more manual dictionary file management. The server automatically syncs its trained dictionary to every connecting client.

• Zero Config: Train a dictionary once on the server with /apex train 120, and every client receives it automatically on connect — no file sharing required.
• Always Fresh: If the server retrains a better dictionary, clients pick up the new version on their next login with full CRC32 integrity verification.

🌟 The Synergistic Engine (Introduced in v1.5)

📦 Unbounded Super Bundling

Forget arbitrary packet limits. ApexNetwork now utilizes Unbounded Micro-Batching. It intelligently accumulates the entire server tick's worth of packets into a single, massive Zstd super-frame (up to 1.5MB).

• Why it's better: Zstd thrives on large, contiguous blocks of data. By compressing 50 packets at once instead of 5, the compression ratio increases exponentially.

🧠 Compression-Aware Traffic Shaping

Traditional limiters choke players based on raw data size. ApexNetwork uses a "Discount" system.

• Smart Tokens: The traffic shaper actively monitors the live compression ratio.
• True Wire Speed: If the server compresses a packet by 80%, the bandwidth cost is discounted by 80%. This allows players to utilize their full internet speed without artificial server-side bottlenecks.

📐 Contiguous Block RLE Deltas

We have completely rewritten the Chunk and Recipe Delta engine.

• Run-Length Encoding: Instead of sending byte-by-byte differences, the engine now tracks contiguous blocks of changes using VarInts.
• Massive Savings: Reduces the size of Delta updates by an additional 90%, making moving through previously loaded areas require almost zero bandwidth.

🛡️ Core Features

• 🗜️ Zstd Compression & NBT Stripping: We replace vanilla Zlib with Zstd — 10x faster compression at comparable ratios, verified by benchmark.
  Experimental: The engine intercepts raw byte streams and replaces common NBT strings with 1-byte IDs before compression begins.
• ⬆️ C2S Movement Quantization: For players with terrible upload speeds.
  Experimental: Instead of sending absolute 64-bit coordinates every tick, the client sends relative 16-bit fractions. Slashes client upload requirements by over 60%.
• 🔄 Flush Consolidation: Inspired by Krypton, this consolidates network flushes. Instead of waking up the network card for every single packet, it groups writes together at the end of the tick, drastically lowering CPU usage.
• 📚 Shared Dictionary Compression: A 112KB trained dictionary captures recurring Minecraft byte patterns (NBT tags, block IDs, entity formats) so even tiny packets under 256 bytes compress efficiently. Automatically synced from server to client — just run /apex train 120 once on the server.
• 📊 Built-in Benchmarking: Run /apex benchmark to generate a comprehensive compression comparison report directly from your server's real traffic data. Compare Zlib, Zstd, Zstd+Dictionary, and LZMA2 across every compression level.

🔧 Compatibility & Usage

Feature	Server Only	Server + Client
Flush Consolidation	✅	✅
Traffic Shaping	✅	✅
Vanilla Client Support	✅	✅
Auto Dictionary Sync	❌	✅
Hybrid Zstd/LZMA2	❌	✅
Asynchronous Bypass	❌	✅
Super Bundling	❌	✅
RLE Chunk Deltas	❌	✅
NBT Stripping	❌	✅
C2S Quantization	❌	✅

📥 Installation Guide

1. Server: Drop the .jar into your mods folder. No configuration changes are required—defaults are pre-optimized.
2. Client (Recommended): Drop the .jar into your mods folder to enable the Asynchronous Engine features and LZMA2 decompression.
3. Train a Dictionary: Run /apex train 120 while actively playing. The server will automatically sync it to all connecting clients.
4. Live Tuning: Use /apex level <level> to dynamically adjust your Zstd or LZMA2 compression limits without restarting. Use /apex benchmark to see detailed performance stats.

❓ Frequently Asked Questions

📜 Permissions & License

• Author: Created by VladiusMissileus.
• Modpacks: You are free to use this in any CurseForge/Modrinth modpack (public or private). No special permission is required!
• Modifications: Modification or redistribution of the code/assets is NOT permitted. This mod is provided as-is for personal use and modpack inclusion only.
• Redistribution: You may NOT repost this mod on other sites (like 9minecraft) or claim it as your own. Please always link back to this official CurseForge page.
• License: All Rights Reserved.

---

Powered by heavily modified Netty pipelines, Zstd JNI, and LZMA2. All performance claims verified via /apex benchmark.