HedgeDB key value store

Hey guys,

I was looking around on LinkedIn (here) and found this project called HedgeDB, which is an embedded key-value store inspired by RocksDB but written with C++20 coroutines, io_uring and O_DIRECT. I am not the creator or anything, I am just trying to understand storage engines and the architectural choices behind them.

The repo shows some crazy benchmarks compared to RocksDB, like 3x or 5x better throughput, but the project is a very early prototype (v0.0.1) and misses basic things like block compression or handling values larger than 4KB. Everytime I see this high performamce metrics I am always skeptical, and never know which path to follow.

As any of you tried it? What's your take on it and where is the catch?

Here is the repo: https://github.com/fede-vaccaro/HedgeDB

https://redd.it/1upzhfz
@r_cpp