A question about performance

Support
1

A question about performance.
In chromadb vector database there is a problem that as the number of records in the storage grows, the time of inserting a new item grows.
I plan to use hybrid search with openAI vectors. And the number of records is planned to be about 2 million. What time of inserting new items will be at such number of records? What search time can be with such number of records ? How many computing resources do you recommend for such a task. Previously used chromadb but in production there are problems with the index at 600k vectors.

2

Hi @IvankoPo

I will say that we are very proud of how Weaviate scales - for example, we can comfortably scale to hundreds of millions of vectors (The Sphere Dataset in Weaviate | Weaviate - vector database) in our index.

While we can’t predict specific times taken for operations as that will be very system-dependent, we do have example benchmarks like these: ANN | Weaviate - vector database

I hope that helps!

Edit: Oh, as far as resource planning goes - we have this page as guidelines :slight_smile: Resource Planning | Weaviate - vector database