Performance Overview
This page summarises the throughput numbers Surgewave currently reaches in its embedded benchmark harness, plus the optimisation areas the codebase is built around. For the full benchmark methodology and per-platform comparison see Benchmarks.
Key Metrics
Comparative latency numbers are preliminary targets and will be published alongside a head-to-head benchmark on identical hardware in the 1.0 release. Throughput numbers below come from the embedded benchmark harness (
benchmarks/baselines/benchmark-baseline.json); see Benchmarks for the full setup.
| Metric | Surgewave Native | Kafka Protocol |
|---|---|---|
| Producer Throughput | 1.25M msg/s | 68K msg/s |
| Consumer Throughput | 1.28M msg/s | 138K msg/s |
Performance Advantages
Native Protocol
- Lower latency than the Kafka wire path (target — public head-to-head benchmark pending)
- Single-pass serialization
- Zero-copy where possible
- SIMD optimizations
Storage
| Backend | Throughput | Latency |
|---|---|---|
| Memory | 625K msg/s | ultra-low (target) |
| ZeroCopyWal | 549K msg/s | ~20 µs |
| File | 286K msg/s | ~50 µs |
Shared Memory
- Sub-microsecond latency
- Lock-free ring buffers
- No network overhead
Optimization Techniques
- SIMD Operations - Vectorized encoding
- ArrayPool - Reduced allocations
- Span/Memory - Zero-copy APIs
- Channel Pipelines - Async batching
- Hardware CRC32 - SSE4.2/ARM acceleration
Next Steps
- Benchmarks - Detailed measurements
- Tuning Guide - Configuration optimization