3 ms·
can vespa index 100M+ vectors on a regular RAM cpu server? any faster w/ a gpu (T4 / A10)?
by lmeyerov 4y ago
can vespa index 100M+ vectors on a regular RAM cpu server? any faster w/ a gpu (T4 / A10)?
- jkb79 4y agoYes, https://blog.vespa.ai/vespa-hybrid-billion-scale-vector-search/ https://blog.vespa.ai/vespa-hybrid-billion-scale-vector-sear...
- lmeyerov 4y agoSuper promising, thanks. Will definitely watch for production readiness - this direction is a big deal as we find most data is cold (logs, docs, ..), not hot, and finally starts becoming more relevant than just doing ES or our own!