Vector storage

Features

RDS for MySQL natively supports vector data processing, including vector storage, similarity calculation, and high-performance index creation. It provides an out-of-the-box vector solution for scenarios such as large-scale semantic search, intelligent recommendations, and multimodal analysis. Use standard SQL interfaces to seamlessly integrate high-precision vector matching with complex business logic. This integration lets you quickly build and deploy innovative AI applications on a low-cost, highly compatible architecture.

• Efficient storage, access, and computation of high-dimensional vectors: Supports storing floating-point vector data with up to 16,383 dimensions and introduces the VECTOR data type. It supports standard SQL interfaces, allowing you to directly write, update, and manage vectorized data in batches. The following table describes the supported vector processing functions.

  Function name	Description
  VECTOR_DIM	Returns the number of dimensions in a vector.
  VEC_FROMTEXT	Converts a string to a vector.
  TO_VECTOR
  STRING_TO_VECTOR
  VEC_TOTEXT	Converts a vector to a string.
  FROM_VECTOR
  VECTOR_TO_STRING
  VEC_DISTANCE	Calculates the distance between two vectors. If one of the operands is an indexed column, the function automatically detects the distance type of the index.
  VEC_DISTANCE_EUCLIDEAN
  VEC_DISTANCE_COSINE

• High-performance vector index: The vector index uses a deeply optimized HNSW (Hierarchical Navigable Small World) algorithm. It uses techniques such as SIMD hardware acceleration, Bloom filter search pruning, and LIMIT condition pushdown to significantly improve retrieval efficiency for large-scale vector data. It also supports hybrid storage and joint queries of both vector data and scalar data.

• Open-source ecosystem and out-of-the-box usability: The feature is fully compatible with the MySQL protocol and supports JDBC/ORM tools and mainstream development frameworks. It integrates with Alibaba Cloud services such as DTS and DMS, providing full-lifecycle capabilities, including data synchronization, management, backup, and recovery. You can upgrade existing instances with a single click without creating a new cluster.

Applicability

• Database version: MySQL 8.0 (minor engine version 20251031 or later). If your instance does not meet the version requirements, you can upgrade the minor engine version or upgrade the major engine version.

• This feature has the following limitations:

  • You can create a vector index only on a table that uses the InnoDB engine.

  • The primary key of the table cannot exceed 256 bytes in length.

  • You cannot use the inplace syntax to create, modify, or delete a vector index.

  • A vector index cannot be set to INVISIBLE.

  • You cannot use the Recycle Bin feature on tables that contain a vector index.

  • Data modification and queries on a vector index support only the Read Committed (RC) isolation level.

  • Due to the randomness of the HNSW algorithm, such as random level assignment and heuristic algorithms, the graph structure of the vector index on the primary/standby instances is not guaranteed to be identical.

  • If a stored procedure or function in the source database uses the vector type, synchronization or migration to a destination database that does not support vectors will fail.

Parameter management

Enable and use vector storage

Step 1: Enable vector support

1. Go to the RDS console, select the target region, and then click the instance ID.

2. On the Basic Information page, in the Running Status section, find Vector Storage and click Enable.

3. When the status changes to Enabled, the feature takes effect immediately.

Step 2: Create a table and a vector index

-- Create a table with a 5-dimension vector column and an HNSW index
CREATE TABLE product_embeddings (
  id INT NOT NULL AUTO_INCREMENT PRIMARY KEY,
  product_name VARCHAR(255),
  embedding VECTOR(5) NOT NULL,
  -- Create a vector index, and specify M (graph connectivity) and the distance metric
  VECTOR INDEX idx_embedding(embedding) M=16 DISTANCE=COSINE
);

Step 3: Insert data

-- Use the VEC_FROMTEXT function to insert vector data
INSERT INTO product_embeddings (product_name, embedding) VALUES
('product_A', VEC_FROMTEXT('[0.1, 0.2, 0.3, 0.4, 0.5]')),
('product_B', VEC_FROMTEXT('[0.6, 0.7, 0.8, 0.9, 1.0]')),
('product_C', VEC_FROMTEXT('[0.11, 0.22, 0.33, 0.44, 0.55]'));

Step 4: Perform a vector similarity search

-- Find the two products that are most similar to the given vector '[0.1, 0.2, 0.3, 0.4, 0.51]'
SELECT
  id,
  product_name,
  VEC_DISTANCE(embedding, VEC_FROMTEXT('[0.1, 0.2, 0.3, 0.4, 0.51]')) AS similarity_score
FROM
  product_embeddings
ORDER BY
  similarity_score ASC -- The smaller the cosine distance, the more similar the vectors are
LIMIT 2;