OSS provides multilayer data protection to ensure durability and reliability. Erasure coding and cross-region replication deliver disaster recovery at the device, data center, and region levels. Versioning and scheduled backups guard against accidental deletion and software bugs. Real-time checksums detect and repair transmission and storage errors, while multi-node redundancy ensures seamless hardware failure recovery.
Zone-redundant storage
OSS uses a multi-zone data redundancy mechanism for regions that have three or more zones. Your data is stored redundantly across at least three zones within the same region. If a zone becomes unavailable, you can still access your data. For regions that have two zones, OSS uses a dual-zone data redundancy mechanism. Your data is stored redundantly across both zones within the same region. If a zone becomes unavailable, you can still access your data.
Supported regions, data durability, and service availability are listed in Storage redundancy.
ZRS incurs higher storage costs than locally redundant storage. For more information, see Object Storage Service pricing.
After you enable ZRS for a bucket, you cannot disable it.
Cross-region replication
Cross-region replication (CRR) asynchronously copies objects across buckets in different OSS regions in near real-time. Destination objects are exact replicas of source objects, preserving names, versions, metadata, content, and ACLs. CRR supports both unencrypted objects and objects encrypted with SSE-KMS or SSE-OSS.
CRR addresses the following use cases:
Compliance: Although OSS stores multiple replicas of each object on physical disks by default, some regulations require data copies at a significant geographic distance. CRR replicates data between distant OSS regions to meet these requirements.
Reduced latency: Maintain object replicas in OSS regions closer to your users to minimize access latency across geographic locations.
Backup and disaster recovery: Store a copy of all data in another region to prepare for major disasters. If one data center is destroyed, activate the backup in the other region.
Data migration: Replicate data between OSS regions for business continuity or migration.
Operational needs: Keep object replicas in multiple regions for compute clusters that analyze the same dataset.
With CRR enabled, Replication Time Control (RTC) replicates most uploaded objects within seconds and 99.99% within 10 minutes. RTC provides near real-time monitoring and replication metrics.
Versioning
OSS versioning protects bucket data from accidental deletion. When you enable versioning, overwritten or deleted objects are saved as previous versions that you can restore at any time.
Versioning applies to all objects in the bucket. Each update generates a unique version ID.
You can upload, list, download, delete, and restore objects in a versioning-enabled bucket.
Suspend versioning to stop creating new versions. Previous versions remain accessible by version ID.
Each version incurs storage charges. Use lifecycle rules to delete expired versions or delete markers and reduce costs.
Scheduled backup
Scheduled backup lets you create automated backup plans with customizable cycles and coverage. Periodic snapshots protect against accidental deletion, logic errors, and disaster events.
Policy customization
Choose backup cycles (hourly, daily, weekly, or monthly) and scope (entire bucket or specific prefix).
Incremental backups reduce storage costs and improve efficiency.
Precise recovery capabilities
Restore data to any previous snapshot with a single click. Supports full restoration or rollback of specific files.
Fast recovery minimizes business downtime.
Retention policy (WORM)
OSS retention policies provide Write Once, Read Many (WORM) protection, preventing objects from being modified or deleted during a specified retention period. Configure a retention policy on a bucket to restrict all users, including resource owners, to read-only and upload access until the retention period expires.