Storage
Cloud Storage
Unified object storage for any amount of data
AWS equivalent
S3
AWS → GCP: Key Differences
- ▸
Bucket names are globally unique (same as S3). But GCS does NOT have S3's concept of path-based regional routing per prefix.
- ▸
Storage classes: Standard, Nearline (30-day min), Coldline (90-day min), Archive (365-day min). Similar to S3 tiers.
- ▸
GCS is the central storage layer for all GCP data services — BigQuery exports, Dataflow staging, Dataproc input/output, ML training data.
Key Concepts to Know
- 1
Objects are immutable. To 'update', you replace the object. Versioning available.
- 2
Lifecycle policies: transition to colder storage class or delete after N days.
- 3
Signed URLs: time-limited access to objects without requiring GCP credentials.
- 4
Notifications: trigger Pub/Sub or Cloud Functions when objects are created/deleted.
- 5
Transfer service: migrate data from AWS S3, Azure Blob, or on-prem to GCS.
DCE Interview Tips
- ★
GCS is the 'data lake' layer for GCP — BigQuery, Dataflow, Dataproc all read/write from GCS.
- ★
For migration: 'We can use Storage Transfer Service to migrate your S3 data to GCS incrementally with zero downtime.'
- ★
Storage class selection: 'Standard for frequently accessed, Nearline for monthly backups, Coldline for DR, Archive for compliance data.'
Common Gotchas
- !
Retrieval costs apply for Nearline, Coldline, Archive — not just storage. Make sure customers understand total cost.
- !
Multi-regional buckets are stored in multiple regions automatically. More expensive.
- !
Object retention locks: cannot delete objects before expiry — useful for compliance.