TECHNOLOGY PROFILE · TECH-0157
Milvus
Distributed vector database.
Reference link only — not reviewed this edition. This profile preserves the prepared catalog’s original descriptions. Advantages depend on the workload; comparisons are editorial, not benchmark results.
Where it has an edge
Distributed indexing and storage target larger multi-collection vector workloads than a simple embedded index.
Best-fit problems
Large-scale vector retrieval requiring a dedicated service.
Tradeoffs & cautions
Deployment complexity can be excessive for a small corpus.
What to test before choosing
Measure recall@k, filtered recall, indexing cost, p95 query latency, updates, backup and access control.
Runtime & hardware
Embedded, self-hosted or managed
Access & licensing
Open-source or commercial depending on edition
Execution
Local or hosted
Lifecycle
Not independently audited
Comparison basis
Editorial problem-fit judgement; no head-to-head benchmark performed
Connected concepts
Alternatives to investigate
Qdrant · Pinecone · FAISS