What hardware do I need to run Qdrant for 1 million vectors?

A 4 vCPU / 8GB RAM server with NVMe SSD handles 1 million 1536-dimensional vectors comfortably when mmap is enabled. Without mmap, budget for 16GB RAM. CPU matters more than RAM for query throughput, with each additional vCPU adding roughly 200 QPS of capacity.

How does Qdrant compare to pgvector for vector search?

pgvector works well for under 100K vectors and teams already running PostgreSQL, but Qdrant is 5-10x faster on query latency beyond that scale with better memory efficiency and purpose-built payload filtering. For new projects targeting more than 100K vectors, choose Qdrant directly rather than adding vector search to a relational database.

Does Qdrant support hybrid search with dense and sparse vectors?

Yes, since v1.10.0 Qdrant can store both dense (neural) and sparse (BM25/TF-IDF) vectors in the same collection and combine them at query time using Reciprocal Rank Fusion (RRF). This combines semantic understanding from dense vectors with exact keyword matching from sparse vectors.

Can I run Qdrant without Docker?

Yes. Qdrant provides pre-built binaries for Linux x86_64, ARM64, macOS, and Windows, downloadable from the GitHub releases page, plus a Homebrew tap for macOS. Docker is still strongly recommended for production due to easier configuration management, volume persistence, and restart policies.

How much RAM does Qdrant use for 1 million vectors with mmap enabled?

With memory-mapping enabled, Qdrant handles 1 million 1536-dimensional vectors using only about 720MB of RAM, delivering 10ms P50 and 22ms P99 query latency. Enabling mmap keeps just the HNSW graph in RAM and memory-maps raw vectors from disk, cutting RAM usage by 60-80% with a performance penalty typically under 15% on NVMe SSD.

Qdrant：由 Rust 驱动的向量数据库，处理 100 万以上向量，延迟 10 毫秒 — 自托管部署指南 2026

Deploy Qdrant vector database for production similarity search. Complete guide to HNSW indexing, payload filtering, multi-tenancy, Docker deployment, and Python/Go/JS clients with real benchmarks.

Apache-2.0
更新于 2026-05-19

Affiliate Disclosure: This article contains affiliate links to DigitalOcean, HTStack, and 虎网云. If you purchase services through these links, dibi8.com may earn a commission at no additional cost to you. All recommendations are based on genuine technical evaluation, not affiliate availability. See our full disclosure policy for details.

Last updated: 2026-05-19. Tested with Qdrant v1.13.0, qdrant-client 1.13.0, Python 3.12.

References & Sources #

Qdrant
Qdrant Documentation
qdrant-client (Python)
Qdrant JavaScript/TypeScript Client
Qdrant Go Client
LangChain
LlamaIndex
Weaviate
Chroma
Milvus
pgvector
Prometheus

📦 出现在以下合集中

📚 知识库 Stack →

References & Sources #

🔗 相关资源推荐

📦 出现在以下合集中

💬 留言讨论