Pgvector vs OpenSearch

Pgvector vs. OpenSearch

Compare Pgvector vs. OpenSearch by the following set of capabilities. We want you to choose the best database for you, even if it’s not us.

Pgvector vs. OpenSearch on Scalability

PgvectorOpenSearch

Separation of storage and compute

Yes. pgvector enables separation of storage and compute by allowing you to store your application data on one database while storing vectors, lookup values, and filter values on a separate database.

❌

Separation of query and insertions

❌

Yes.

Multi-replication

✔️

❌

Dynamic segment placement vs. static data sharding

❌

Both

Cloud-native

❌

Billion-scale vector support

❌

pgvector scalability

You can use a solution like YugaByteDB to extend the capabilities of Postgres for distributed environments.

OpenSearch

OpenSearch supports horizontal scaling, cluster management optimizations, and efficient shard allocation, making it suitable for handling large datasets and high query loads effectively.

Pgvector vs. OpenSearch on Functionality

Performance is the biggest challenge with vector databases as the number of unstructured data elements stored in a vector database grows into hundreds of millions or billions, and horizontal scaling across multiple nodes becomes paramount.

Furthermore, differences in insert rate, query rate, and underlying hardware may result in different application needs, making overall system tunability a mandatory feature for vector databases.

PgvectorOpenSearch

Role-based Access Control (RBAC)

✔️

❌

Disk Index support

❌

Hybrid Search (ie Scalar filtering)

Yes. Sparse & Dense Vectors and Scalar filtering.

yes, vector search & keyword search & scalar filtered search

Partitions/namespaces/logical groups

❌

Index type supported

HNSW & IVFFlat

ANN

Multiple in-memory indexes support

❌

OpenSearch

OpenSearch supports:

Vectors with up to 16,000 dimensions.
Vector generation through external libraries or directly within OpenSearch.
Both binary and dense vectors.
Cosine Similarity, Inner Product, and L2 Distance (Euclidean).
Integration with multiple engines, including NMSLIB, Faiss, and Lucene, to facilitate vector indexing and searching.

Pgvector vs. OpenSearch on Purpose-built

PgvectorOpenSearch

Purpose-built for Vectors

pgvector is an add-on to Postgres

No. Vector search is an add-on to OpenSearch.

Tunable consistency

✔️

❌

Support for both stream and batch of vector data

✔️

❌

Binary Vector support

❌

Multi-language SDK

Use pgvector from any language with a Postgres client

Java, Python, JavaScript, Go, and .Net

Database rollback

✔️

❌

Pgvector vs. OpenSearch: what’s right for me?

Pgvector

pgvector is a PostgreSQL extension designed to facilitate the storage, querying, and indexing of vectors within a PostgreSQL database.

License: PostgreSQL License

OpenSearch

OpenSearch is an open-source software suite for search, analytics, security monitoring, and observability applications. It is not purpose-built for vector storage and search workloads but introduces a vector search plugin to provide this capability. Amazon OpenSearch Service is an AWS-managed service for OpenSearch.

Apache 2.0