Building now

GooseDB

A full database system I'm building from the ground up in C++23, for storing and querying media.

GooseDB is meant to work more like a distributed, S3-style store than a classic single-machine database: data spread and replicated across machines, with a query layer on top.

I'm building it from the ground up as I work through Carnegie Mellon's database course, starting with the storage engine: 4 KB pages that hold tuples, each with a header, and table schemas that describe their columns.

Clients

Store and query media

Query layer

Parse, plan, execute

Storage engine

4 KB pages of tuples, schemas

Many machines

Spread and replicate data

The plan, simplified. The storage engine is being built now; dashed boxes are still to come.

See it work

Where it stands: a page, from include/storage/page.hpp

static constexpr std::size_t PAGE_SIZE = 4096;

class Page {
public:
    Page(Header header) : header(header) {}

    void add_tuple(const std::vector<std::byte>& d) {
        std::size_t free_offset = header.get_free_space_offset();
        std::size_t tuple_count = header.get_tuple_count();
        header.set_tuple_count(tuple_count + 1);
        std::copy(d.begin(), d.end(), data.begin() + free_offset);
    }

private:
    Header header;
    std::vector<std::byte> data = std::vector<std::byte>(PAGE_SIZE);
};

Every page is 4 KB. Adding a tuple bumps the count in the page's header and copies the bytes in at the free-space offset.