Storage: OSS, NAS, Block Storage, Hybrid Backup
Match storage to access pattern
- Block storage (cloud disks) - attached to one ECS instance, low latency, for OS and databases. Types: ESSD (high performance, PL levels), SSD, ultra disk, basic disk. Snapshots provide point-in-time backup.
- Object storage (OSS) - HTTP-accessible, unlimited scale, for unstructured data, backups, static assets. Storage classes: Standard, Infrequent Access, Archive, Cold Archive; lifecycle rules transition objects between classes. Buckets are regional and named globally unique.
- File storage (NAS) - shared POSIX file system mountable by many ECS instances simultaneously; good for shared content and lift-and-shift file workloads. Capacity and performance scale types differ.
Data protection
Hybrid Backup Recovery (HBR) centralizes backup of ECS, OSS, NAS, and on-premises data. OSS versioning and cross-region replication protect against overwrite and region loss. Snapshot policies automate disk backup.
Traps
- Assuming OSS is a file system - it is object storage with eventual-consistency nuances for overwrite listing; not a drop-in for POSIX apps.
- Forgetting that a cloud disk attaches to a single instance (except shared block devices) - multi-instance sharing needs NAS.
- Assuming Archive-class objects are immediately readable - they require restore first, which takes time and incurs cost.
- Confusing snapshots (disk-level) with HBR backup (policy-driven, cross-resource).
- Ignoring lifecycle rules that silently move hot data to cold tiers and slow access.
Rule of thumb
One writer, low latency → block. Many readers/writers needing shared files → NAS. Web-served or archival unstructured data → OSS with lifecycle rules. Always pair with a backup or replication strategy; storage redundancy is not backup.
Worked example
A media company stores originals in OSS Standard, transitions to IA after 30 days and Archive after 180 days via lifecycle rules, replicates the bucket cross-region, and uses HBR for NAS-backed editing shares.