Design of a Generic Backup and Disaster Recovery Solution

In recommendation scenarios, strict SLA requirements and downstream inconsistencies often necessitate backup and disaster recovery (DR) mechanisms.

The proposed solution provides a universal backup‑DR framework supporting pagination and bucketed data, allowing multiple pages and buckets per scenario to avoid large‑scale data inconsistencies during failures.

Overall Design

The system architecture consists of three business layers: backup data, DR retrieval, and dirty‑data cleanup. Backup can be manual or scheduled; cleanup can be automatic or manual. Underlying services handle data storage, backup methods, and logging.

Backup Data

Automatic backup uses scheduled tasks to fetch service data; manual backup relies on historical user data. When data volume is large, sharding (implemented by the business) is supported.

Dirty Data Cleanup

Automatic cleanup removes time‑sensitive data via scheduled jobs; manual cleanup targets identified dirty records.

DR Process

DR configuration includes downgrade flags, scene name, bucket settings, filter chain, and backup strategy selection. Handlers may implement exposure deduplication, sharding assembly, and time‑based filtering.

The design offers proactive backup, customizable exposure filtering, bucketed delivery, sharding support, and both manual and automatic dirty‑data cleanup.