Every relational and document database—whether PostgreSQL, MySQL InnoDB, SQLite, or TiKV—relies on an underlying storage engine to organize bytes on persistent storage, maintain memory caches, and survive sudden power failures without data corruption.
While application developers interact with SQL queries and transactions, the storage engine operates in a world of disk blocks, slotted page headers, buffer pool frame tables, and sequential write-ahead log files. The performance and durability of a database depend on how effectively its storage engine bridges the massive latency gap between nanosecond CPU cache operations, microsecond RAM access, and millisecond disk I/O.
This module deconstructs the internal mechanics of database storage engines: physical slotted page formats, memory buffer pool managers and replacement policies (LRU-K / 2Q / Clock), concurrent B+ Tree latch crabbing, Write-Ahead Logging (WAL) rules, and the classic ARIES crash recovery algorithm.