IPA IP 2026 Question 56

Writing the updates held in memory to an external storage device all at once, or the moment this is done — a checkpoint shortens recovery after a failure.

For speed, a DBMS keeps updated data in buffers in memory and writes it to the database on disk later. At a checkpoint it writes all the updates made so far by the transactions to the external storage device, such as an HDD, at one time, and records that point in the log. After a system failure, recovery only needs to start from the last checkpoint: transactions committed after it are redone from the log (roll forward), and those not committed at the time of the failure are undone (rollback). Without checkpoints, recovery would have to replay the log from much further back.

Answer (d)

Why not others:

• •(a) — making a transaction’s processing final when it ends normally is a commit
• •(b) — returning the database to its state before the transaction began when it fails is a rollback
• •(c) — transactions each waiting for the other to release its lock, so that neither can go on, is a deadlock

Key rule: Commit confirms a transaction, rollback undoes it, a deadlock is two transactions waiting on each other’s locks, and a checkpoint writes the memory buffers to disk so that recovery can start from there.

Databases · AI-written, may contain errors