Are there any benchmark results for the memory-constrained setting?

cmuparlay / ParlayANN

A library of algorithms for approximate nearest neighbor search in high dimensions, along with a set of useful tools for designing such algorithms.

MIT License

88 stars 21 forks source link

Are there any benchmark results for the memory-constrained setting? #16

Closed s-jse closed 3 months ago

s-jse commented 3 months ago

Hi,

I was wondering if you have tested your implementation on a machine with limited RAM, where most of the index has to be kept on disk? Especially for DiskANN, which was originally designed to work well in this setting.

magdalendobson commented 3 months ago

Hi! Currently the library supports in-memory search only and we don't have an SSD-optimized graph search routine. If you do any benchmarking of this manner using ParlayANN, we would love to hear about it.