The book suggests building a database on top of a KV store. This is precisely what I did with my project. I was initially building a product that I hoped would

The book suggests building a database on top of a KV store. This is precisely what I did with my project. I was initially building a product that I hoped would be able to replace existing file systems and I needed an architecture that made it easy to create meta-data tags for every file and then find every file that had certain tags very quickly. I implemented the tags using a set of novel Key-Value stores that I invented. Once I had it working, I realized the KV stores I used for tags were just like columns in a relational table built within a columnar database. Querying for files based off their tags was very much just like SQL queries for table rows. So I tried using them to create relational tables. They turned out to be incredibly fast at a variety of queries (the bread and butter of databases) without needing to create separate indexes in order to get optimal performance. I thought database experts would be intrigued when I showed how much faster my system was than other conventional RDBMS setups on the same hardware. I guess I was surprised when almost no one was even curious how it did it. Here is a simple, short video comparing it to SQLite: https://www.youtube.com/watch?v=Va5ZqfwQXWI https://www.youtube.com/watch?v=Va5ZqfwQXWI