Testing on 5GB of data fully resident in ram is a terrible comparison. Things get hard when you're in the hundreds of gigabytes or more.

Thanks a lot for your comment! We agree that a dataset as small as 5 GB may sound strange but it was a conscious decision. Check out our blog post to read more about the methodology of this benchmark itself.

https://blog.serenedb.com/search-benchmark-game-overview

TLDR It's not our choice, but it's meaningful. Because this 5GB is single data segment and literally what you will have in Elastic/etc when you have overall TBs of data. See https://www.elastic.co/docs/deploy-manage/production-guidanc... (single shard is one Lucene index that contains multiple data segments)

Great results! Refreshing to see a project that actually went the extra mile and built the core search engine in C++ from scratch, unlike most similar projects that just wrap an existing library.

Thank you for your feedback! We have long thought that the space was missing an alternative.

Very impressive numbers! I def would love now to learn more about the algo behind these results

Thanks! There some articles on our blog. Check it out!

I can't wait to apply this base to my real-time AI forecasting solutions!

Hi Leonid! More than happy to talk about this! Find me on social networks by by nickname.