🧷Checkpoint

NEW

A saved system or process state recorded periodically to enable faster recovery or restart after failures.

DESCRIPTION

Checkpoints persist the current state of a system—such as database buffers, metadata, or application memory—to durable storage. After a crash, recovery starts from the latest checkpoint and replays logs to reach consistency, reducing restart time. In databases, checkpoints bound redo work and control write bursts. In stream processing, checkpoints capture operator state to provide exactly-once semantics. Tuning checkpoint frequency balances overhead and recovery time.

ALTERNATIVE DEFINITIONS

Wikipedia

A checkpoint is a saved state of a process or system from which it can be restarted in the event of a failure.

General computing definition emphasizing restart from saved state.

Source

PostgreSQL

Checkpoints are points in the WAL sequence at which all data files have been updated to include the information logged before that point.

Database-specific definition linking checkpoints to WAL.

Source

RELATED TERMS

Write-Ahead Logging ARIES Recovery Transaction Log Rollback Incremental Backup

COMPETING TERMS

Materialized View

KNOWLEDGE AREAS

Data Storage & Operations Infrastructure Observability & Telemetry Software Engineering

TARGET AUDIENCE

Data Engineer DevOps Engineer Software Developer IT Manager Data Architect

EXTERNAL SOURCES

• Wikipedia - Checkpoint (computing)
• PostgreSQL - Checkpoint and WAL configuration
• Apache Flink - Flink streaming checkpoints
• Apache Spark - Spark streaming checkpointing

TOP QUESTIONS

Why are Checkpoints used in databases?

They bound recovery work by ensuring changes up to a point are flushed, reducing restart time after crashes.

Category: Relevancy

How do Checkpoints work with write-ahead logging?

WAL records changes first; checkpoints later persist dirty pages so redo only applies after the checkpoint.

Category: Example

What is the trade-off in checkpoint frequency?

Frequent checkpoints increase runtime overhead but shorten recovery; infrequent checkpoints reduce overhead but lengthen recovery.

Category: Risks

How do streaming systems use Checkpoints?

They capture operator state to enable exactly-once or at-least-once processing upon failures.

Category: Example

Are Snapshots the same as Checkpoints?

Snapshots often capture complete state images; checkpoints may be incremental or log-coordinated. They serve similar recovery goals with different costs.

Category: Alternatives

What is a common mistake with Checkpoints?

Poorly sizing I/O or buffers, causing checkpoint spikes and latency in transactional workloads.

Category: Mistakes

How do Checkpoints affect durability guarantees?

They reinforce durability by periodically syncing dirty state to persistent storage, complementing logging.

Category: Example

What metrics indicate healthy Checkpointing?

Checkpoint duration, frequency, write throughput, and recovery time objectives (RTO) are key indicators.

Category: Management

Can Checkpoints be coordinated across services?

Yes, but coordination adds complexity; many systems favor independent checkpoints with idempotent recovery.

Category: Opinion