The short answer
stores files as objects in buckets (S3, GCS, ADLS/Blob). You address them by key over HTTP APIs. Ideal for data lakes, backups, and shared analytics files.
Block storage is a disk volume attached to a VM (EBS, Persistent Disk, Azure Disk). The OS sees blocks/filesystems. Ideal for database data directories and boot disks.
Object: app --HTTP--> s3://bucket/file.parquet (many readers)
Block: VM --attach--> /dev/xvdf mounted at /data (one instance*)\*Some cloud disks can multi-attach in special cases; default mental model is one VM.
Comparison
| Object | Block | |
|---|---|---|
| Interface | Object API / keys | Disk / filesystem |
| Sharing | Easy across services | Tied to compute (usually) |
| Use in DE | Lake files, Parquet, logs | DB volumes, local disks |
| Metadata | Rich object metadata | FS metadata |
Interview tip: Lakes use object storage; databases usually sit on block volumes. Spark reads Parquet from object storage, but executors may still use local block disks for shuffle spill.