NVIDIA / NVIDIA/cuvs

[FEA] Out-of-core search

Open
#2,308 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

feature request
Dominant language
Cuda
Stars
854
Forks
236
Avg merge
3d 3h
Merged PRs (30d)
62

Description

Ability to search efficiently on a commodity system (including cloud instances) without requiring the entire index fit in GPU memory.

This could be accomplished based on existing work, but may also require additional work.

The goal here is to achieve better price-perf and raw perf by enabling large-scale indexing through scale-up. This particular issue assumes we still have enough host memory to store the index, so this does not involve disk yet.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reviewing the linked GPU-Graph-ANN work and the cuVS search implementation to identify the relevant entry points for indexes that exceed GPU memory. The work is done when large-scale indexing and search use host memory without requiring disk, while improving the stated price-performance or raw-performance goals.

Written by the indexing model from the issue text.

Assessment

Domain
search
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Quiet
Clarity
Needs clarification
Newbie friendliness
30/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.