Evaluation & Benchmarks
HyperspaceDB is optimized for two critical metrics: Throughput (Ingestion speed) and Latency (Search speed).
Test Environment
- Hardware: Apple M4 Pro (Emulated Environment) / Linux AVX2
- Dataset: 1,000,000 vectors, 1024 Dimensions, Random Distribution in Unit Ball.
- Config:
ef_construction=400,ef_search=400
Results
🚀 Ingestion Speed
Thanks to the Async Write Buffer (WAL) and background indexing, ingestion does not block user requests.
| Count | Time | Throughput | Storage Sements |
|---|---|---|---|
| 10,000 | 0.6s | 15,624 vec/s | 1 |
| 100,000 | 6.5s | 15,300 vec/s | 2 |
| 1,000,000 | 64.8s | 15,420 vec/s | 15 |
🔍 Search Latency (1M Scale)
At 1 million vectors, search performance degrades linearly with graph depth ($\log N$), proving effective HNSW implementation.
| Metric | Value |
|---|---|
| QPS | 14,668 queries/sec |
| Avg Latency | 0.07 ms |
| P99 Latency | < 1.0 ms |
Why is it so fast?
- ScalarI8 Quantization: Fits 8x more vectors in CPU cache.
- No
acosh: Inner loop uses a monotonic proxy function ($\delta$). - SIMD: Vector operations use platform-specific intrinsics.