ARIES is a robust recovery algorithm based on the Write-Ahead Logging (WAL) protocol, widely used in modern DBMS to ensure durability and consistency after system failures.
Types of Log Records in ARIES
Every update operation creates a log record, which can be of the following types:
- Undo-only Log Record: Stores only the before image (original data). Used to undo changes.
- Redo-only Log Record: Stores only the after image (new data). Used to redo changes.
- Undo-Redo Log Record: Stores both before and after images. Used for both undo and redo operations.
Log Sequence Numbers (LSN)
- Every log record is assigned a unique, monotonically increasing LSN.
- Every data page has a pageLSN indicating the LSN of the last update applied to it.
Write-Ahead Logging (WAL) Rules
- A log record must reach stable storage before the corresponding data page is written to disk.
- Log writes are buffered in a memory area called the log tail.
- The log tail is flushed to disk when full.
- A transaction is not considered committed until its commit record is flushed to disk.
Checkpoints
Periodically, the system writes a checkpoint record to the log.
A checkpoint contains:
- Transaction Table
- Dirty Page Table
A master log record, stored in stable storage, holds the LSN of the most recent checkpoint.
Recovery Phases in ARIES
When the system restarts after a crash, ARIES performs three recovery phases:
1. Analysis Phase
- Starts from the latest checkpoint.
- Reconstructs the state of active transactions and dirty pages at the time of the crash.
2. Redo Phase
- Replays operations forward from the smallest LSN of dirty pages.
- Reapplies necessary changes to bring the database to the state before crash.
3. Undo Phase
- Scans the log backward.
- Rolls back the effects of uncommitted (loser) transactions.
