Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

memory-map/on-disk index support #758

Closed
n0thing233 opened this issue Feb 9, 2023 · 3 comments
Closed

memory-map/on-disk index support #758

n0thing233 opened this issue Feb 9, 2023 · 3 comments

Comments

@n0thing233
Copy link

n0thing233 commented Feb 9, 2023

Is your feature request related to a problem?
With limited memory resource, it is expensive to load the whole index file to memory and do the knn search

What solution would you like?
The Faiss backend natively support loading on-disk index in a memory-mapped way and do the search on the fly:
https://github.com/facebookresearch/faiss/wiki/Indexes-that-do-not-fit-in-RAM
I would like it to be implemented with open search , or is it already implemented?

What alternatives have you considered?
N/A

Do you have any additional context?
N/A

@vamshin vamshin added the question Further information is requested label Feb 10, 2023
@jmazanec15
Copy link
Member

Hi @n0thing233, in general, I think leveraging disk based ANN approaches could be very beneficial.

There are a couple issues with supporting faiss's implementation:

  1. We wouldnt be able to support it on all platforms. I dont believe it works on Windows
  2. It requires that several smaller indices be merged together to form the on disk based index: https://github.com/facebookresearch/faiss/wiki/Indexes-that-do-not-fit-in-RAM#building-an-on-disk-index. This would be difficult to do for every segment file, unless we decided to create multiple faiss index files per segment.

All that being said, it is definitely worth looking into supporting some form of disk based ANN.

@navneet1v navneet1v added backlog and removed question Further information is requested labels Mar 27, 2023
@vamshin
Copy link
Member

vamshin commented Mar 28, 2023

Please +1 if you are looking for this feature to help prioritize

@jmazanec15
Copy link
Member

Lets track in #1134

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

4 participants