This directory contains benchmarking tools and results for the front service performance testing.

• front-benchmark/ - Go-based HTTP load testing tool
• run_benchmark.sh - Automated benchmark script
• results/ - Benchmark results storage

• Brotli + Gzip compression - dual compression for all assets
• Code splitting - vendors separated by usage (editor, charts, datetime)
• Emoji data split - 30 language-specific chunks, lazy loaded
• Translation grouping - 12 language bundles instead of 697 JSON files
• Chunk size optimization - minSize: 50KB to avoid tiny files

• In-memory file cache - pre-loaded with pre-computed headers
• Last-Modified caching - single stat() at startup
• HTTP/2 optimized chunks - max 244KB for multiplexing
• Brotli priority - serves .br files when client supports

• Vendors: 58MB → 10.4MB (-82%)
• Vendors (Brotli): 2.3MB (-79%)
• JS files: 1,030 → 110 (-89%)
• Total files: 2,900 → 594 (-80%)

# Run benchmark against current front service
cd front-benchmark
go build -o front-benchmark .

# Benchmark with random files from container
./front-benchmark \
  -url http://localhost:8087 \
  -c 50 \
  -d 30s \
  -random \
  -container dev-front-1

# Benchmark specific URL
./front-benchmark \
  -url http://localhost:8087/config.json \
  -c 50 \
  -d 30s \
  -exact

# Run full benchmark suite
./run_benchmark.sh

See results/ directory for detailed benchmark reports.

Changes:

• Streaming for large files - res.end(buffer) replaced with Readable.pipe(res) for files > SEND_BUFFER_SIZE (default 64KB). Prevents event loop blocking during large file transfers.
• Selective middleware - body-parser and express-fileupload removed from global middleware chain, applied only to POST routes that need them.
• Configurable buffer size - SEND_BUFFER_SIZE env variable to tune streaming threshold.
• Mixed workload benchmark - new -mixed flag to test config.json latency during concurrent file loading.

Key findings:

• Node.js with streaming: config.json stays responsive (20ms avg) under heavy file load
• Nginx proxy_pass adds latency for config.json in mixed workload (59ms avg)
• Nginx uses less memory (~170MB vs ~350MB) since files served from filesystem
• Node.js is 2x faster for API endpoints (config.json) due to no proxy overhead

After optimizations (Brotli/Gzip compression, improved caching, code splitting):

Key Improvements (Node.js vs previous):

• config.json: +75% RPS (6,172 → 10,790)
• Static files: +7% throughput (507 → 541 MB/s)
• Memory: Stable (~260-340 MB)

Conclusion:

• Node.js now significantly faster for API endpoints (+122% vs Nginx)
• Nginx still faster for static files (+20% RPS)
• Both implementations viable depending on workload

-url string        Base URL to benchmark (default "http://localhost:8087")
-c int             Number of concurrent connections (default 50)
-d duration        Benchmark duration (default 30s)
-t duration        Request timeout (default 10s)
-files string      File with list of paths to request
-random            Use random files from container
-container string  Docker container name (default "dev-front-1")
-exact             Use URL exactly as provided
-monitor-memory    Monitor container memory usage
-mixed string      URL to request concurrently with files (e.g., /config.json)
-mixed-conns int   Connections dedicated to mixed URL (default 10)