hanhossain / rdb

rdb

A database written in rust

Design

Storage

Page Cache (Buffer manager)

LRU cache to manage the pages in memory. When a page gets evicted, it will ask the storage layer to flush the page. This also supports flushing all pages (used for periodic flushing).

File Manager

Directly interacts with the file system. When the page manager requests a page from storage, the file manager will handle opening, reading, and writing to the file.

Disk Representation

B+ Tree

• Node types
  • Inner Node
    • Number of max keys in a node depends on key size.
      • Won't vary in a tree, can be part of tree metadata.
  • Leaf Node
    • Number of max key value pairs will vary depending on the size of the key + value.
    • primary tree will have all tuples, index trees will only have the index and primary key.

    Layout
    |----------------------------------------------------------------|
    | |-Header-||-Node Header-||-Leaf Header-----|                   |
    | |  size  ||  node type  || previous | next |                   |
    | |--------||-------------||-----------------|                   |
    | |-Tuple0----------------------||-Tuple1----------------------| |
    | | column0 | column1 | column2 || column0 | column1 | column2 | |
    | |-----------------------------||-----------------------------| |
    |--------------------------------------------------------------- |
    

• node buffer is a full page

KV Store

A key-value store for system metadata. Supports

• insert/update value
  • Uses an append-only process. The store will use the newest object in the blob.
  • Eventually this should have a compaction process to clean up the old objects.
• get value by key

Tables

System Defined

• Holds metadata about the system.
• Will be queryable but will not allow mutable statements.
• Queries will be recursive
  • all queries will reach out to the system table to get metadata
  • system metadata will be hardcoded as the base case.

Incomplete breakdown of what we need to store

users:
  columns:
    id: int
    username: string (20)
    firstName: string (20)
    lastName: string (20)
  primaryKeys:
    - id

User Defined

Example table

id	username	firstName	lastName
1	ironman	Tony	Stark
2	hulk	Bruce	Banner

TODO:

• BTree
  • Iterator without filter
  • Iterator with filter
  • Insert with leaf node split
  • Insert with inner node split
  • Update tuples
  • Remove
  • Compound primary key
• Sql implementation