Durability is almost entirely implemented using WAL (Write-Ahead Logging) in most modern databases.

✅ 1. DURABILITY — How DB ensures data is never lost

Durability means:

Once a transaction is committed → data will never be lost even if the server crashes immediately.

To achieve this, databases do not trust RAM (because RAM is volatile). So they use a persistent mechanism: WAL logs.

⭐ 2. Yes — WAL (Write-Ahead Log) is the main mechanism

Before writing the actual data to data files, the DB first writes the change to the WAL.

WAL contains:

Why “write-ahead”?

Because the WAL is written before changes are applied to the main data files.

⭐ 3. Crash happens — how recovery works

Event:

You book a ticket → DB commits → crash occurs.

After restart, DB does:

🔄 ROLL FORWARD (Redo Phase)

DB replays committed WAL entries to restore all committed data.

🔄 ROLL BACK (Undo Phase)

DB undoes transactions that were not committed.

As a result:

✔ All committed data stays ❌ All uncommitted data is discarded

⭐ 4. Example flow

Step 1: You run:

UPDATE account SET balance = balance - 100;

Step 2: DB writes to WAL:

T100: old_balance=500, new_balance=400
T100: COMMIT

Step 3: DB crashes BEFORE writing to final data file.

💥 Crash.

Step 4: DB restarts

It reads WAL → sees COMMITTED entry → applies the change to the data file.

✔ Your booking / update still exists → durability upheld.

⭐ 5. Why does WAL guarantee durability?

Because WAL is stored on disk, not memory. Even if:

→ WAL file is already safely persisted.

WAL is append-only, so it’s extremely fast and safe to write.

⭐ 6. All major databases use WAL

DB Durability mechanism
PostgreSQL WAL
MySQL (InnoDB) redo logs + undo logs
Oracle redo logs + undo + archive logs
SQL Server transaction log
MongoDB journal / oplog
Cassandra commit log

All follow the same principle: Commit → write log → guarantee durability → recover using logs.

⭐ 7. Final Summary

You asked: “If server crashes, how will data recover? From WAL logs?”

✔ Yes ✔ WAL is the main mechanism ✔ On restart DB replays WAL to restore all committed transactions ✔ Therefore, durability is guaranteed even after sudden failure