Reliability for Database Management Systemss
Database reliability covers how a DBMS survives crashes and, in distributed systems, machine failures. Write-ahead logging records changes before pages reach disk, and the ARIES algorithm uses analysis, redo and undo phases to recover. Distributed databases add replication, partitioning and commit protocols such as two-phase commit.
Stage: Transactions and Distribution · Part of the Database Management Systems roadmap · Last reviewed 19 November 2026
What to learn in Reliability
- Explain the write-ahead logging rule and why log records are flushed first
- Describe the analysis, redo and undo phases of ARIES recovery
- Explain checkpoints and how they shorten recovery time
- Compare steal and no-force buffer policies
- Describe two-phase commit and its blocking problem
- Compare partitioning and replication strategies in distributed databases
Reliability interview questions
What is write-ahead logging?
Write-ahead logging requires that the log record describing a change reaches stable storage before the changed data page does, and before commit returns. After a crash, the log lets the system redo committed work and undo unfinished work.
Why can two-phase commit block?
If the coordinator fails after participants vote yes but before sending the decision, participants cannot safely commit or abort and must wait, holding locks. Consensus-based protocols or three-phase variants reduce this, at extra cost.
Where Reliability fits in the Database Management Systems roadmap
Reliability belongs to the Transactions and Distribution stage (advanced level). Correctness at scale.
Other units in this stage
Back to the full Database Management Systems roadmap