Performance

Benchmarked on real workloads. Designed for production throughput.

Search Performance

Full-text search (10K docs)< 1msP99 latency
Full-text search (1M docs)< 5msP99 latency
Semantic vector search< 10msP99 with 768-dim vectors
Hybrid search (text + vector)< 15msP99 combined
Concurrent queries10K+ QPSOn commodity hardware

Ingestion Performance

Text document ingestion10K docs/sSustained throughput
Batch ingestion API50K docs/sWith batch sizes of 1000
Image processing pipeline36 img/sGPU-accelerated (RTX 4090)
Audio transcriptionReal-time1x speed with Whisper
PDF extraction100 pages/sText extraction

Scalability

Single node capacity10M+ docsWith NVMe storage
Cluster capacityUnlimitedHorizontal scaling
Index size efficiency~30%Of raw data size
Memory usage< 2GBBaseline for 1M docs
Startup time< 3sCold start with warm cache

Run Your Own Benchmarks

Start a 30-day trial and test DataFuse against your own data and workloads.

Start Free Trial