I've been learning how various database components work (e.g., log-structured merge trees) by implementing them from scratch and benchmarking their performance with different workloads and configurations. Sharing the results at https://dbfromzero.com