Performance Benchmarks

Benchmark suite for durable-streams-java using the official upstream @durable-streams/benchmarks.

Prerequisites

• JDK 21+ (for Virtual Threads optimal performance)
• Node.js 18+
• Quarkus server running

First time setup:

cd benchmarks
npm install

Storage Backends

InMemoryStreamStore (Default)

Fast in-memory storage for testing and benchmarking raw protocol performance.

Start server:

cd ..
./gradlew :example-quarkus:quarkusRun

Run benchmarks:

cd benchmarks
npm install
npm run bench:inmemory

BlockingFileStreamStore

Persistent file-based storage using Virtual Threads + blocking I/O.

Start server:

cd ..
./gradlew :example-quarkus:quarkusRun --args='-Dquarkus.profile=filestore'

Run benchmarks:

cd benchmarks
npm run bench:filestore

Benchmark Metrics

The upstream benchmark suite measures:

1. Latency (Round-Trip Time)

• Baseline Ping: Network latency only
• Total RTT: Full append + long-poll cycle
• Overhead: Durable Streams protocol overhead

Success Criteria: < 10ms overhead

2. Message Throughput

• Small Messages (100 bytes): Target 100+ msg/sec
• Large Messages (1MB): Measure with high concurrency

3. Byte Throughput

• Streaming: Target 100+ MB/sec

Results

Results are saved to benchmark-results.json with statistics (min/max/mean/p50/p75/p99).

Example output:

=== BENCHMARK RESULTS ===
Environment: quarkus-inmemory
Base URL: http://127.0.0.1:4432

┌──────────────────────────┬────────────┬────────────┬────────────┬────────────┬────────────┬────────────┬────────────┐
│ (index)                  │ Min        │ Max        │ Mean       │ P50        │ P75        │ P99        │ Iterations │
├──────────────────────────┼────────────┼────────────┼────────────┼────────────┼────────────┼────────────┼────────────┤
│ Latency - Total RTT      │ '5.23 ms'  │ '12.45 ms' │ '7.82 ms'  │ '7.50 ms'  │ '8.90 ms'  │ '11.20 ms' │ 10         │
│ Latency - Overhead       │ '2.10 ms'  │ '8.30 ms'  │ '4.15 ms'  │ '3.80 ms'  │ '5.20 ms'  │ '7.50 ms'  │ 10         │
│ Throughput - Small       │ '350 msg/s'│ '520 msg/s'│ '425 msg/s'│ '420 msg/s'│ '480 msg/s'│ '510 msg/s'│ 3          │
│ Throughput - Large       │ '12 msg/s' │ '18 msg/s' │ '15 msg/s' │ '15 msg/s' │ '17 msg/s' │ '18 msg/s' │ 2          │
└──────────────────────────┴────────────┴────────────┴────────────┴────────────┴────────────┴────────────┴────────────┘

Comparing Implementations

Run both storage backends and compare results:

npm run bench:all

This runs:

1. bench:inmemory - Tests raw protocol performance
2. bench:filestore - Tests real-world persistent storage performance

Compare the two benchmark-results.json outputs to see the performance difference between in-memory and file-based storage.

Custom Configuration

Override benchmark settings via environment variables:

BASE_URL=http://localhost:8080 ENVIRONMENT=my-custom-env npm run bench:inmemory

Troubleshooting

Server not responding?

• Ensure server is running on correct port (default: 4432)
• Check curl http://127.0.0.1:4432/health (add health endpoint if needed)

Benchmarks timing out?

• Increase Vitest timeout in test files
• Reduce concurrency/message counts for slower systems