Why Vectors Need a Dedicated Database and How Milvus Solves It

What Is a Vector?

A vector is a sequence of numbers that encodes the features of an object, e.g., an image, a piece of text, or a person. For a person, gender, age, height and weight form a 4‑dimensional vector; adding education, occupation, interests, etc., can expand the vector to dozens or hundreds of dimensions.

Why Vectors Matter in Modern AI

In contemporary AI workloads, a sentence may be transformed into a 768‑dimensional vector, an image into a 1024‑dimensional vector, and audio into similarly high‑dimensional representations. Storing such vectors in traditional relational databases (MySQL, Oracle) is impractical because they cannot efficiently handle high‑dimensional data or similarity‑based queries.

select * from student where id = 100;

select * from student where name like '%张三%' order by id desc;

select class, count(*) as student_count from student group by class;

Attempting to store vectors as JSON strings in a relational table quickly leads to billions of numeric entries and slow brute‑force similarity calculations.

Introducing Milvus: A Vector‑Specialized Database

Milvus is an open‑source vector database built to store massive high‑dimensional vectors, create indexes (HNSW, IVF_FLAT, FLAT), and return the most similar vectors within milliseconds, even at scales of millions to billions of vectors.

Typical Milvus Workflow

The usage pattern is fixed:

Deploy Milvus.

Create a collection and define the vector dimension.

Insert vectors (and optional payload).

Build an index (e.g., HNSW).

Perform similarity search to retrieve the top‑N nearest vectors.

Manage and delete data as needed.

Common AI Scenarios Powered by Milvus

Typical applications include semantic sentence search, image similarity retrieval, and personalized product recommendation—any task that requires finding the top‑N vectors most similar to a query vector.

Summary

Vectors digitize the characteristics of entities; a vector database stores these vectors and enables fast similarity search. Milvus is currently the most widely adopted, production‑grade open‑source vector database.