What is Object Storage — Cloud Storage in Google Cloud?

2 / 24

Google Cloud

Object Storage — Cloud Storage

Store objects in buckets with one API across storage classes and global/regional location.

Cloud Storage holds objects in buckets, addressed by name over HTTP. There are no real folders.

A storage company with one membership card that opens every facility from the busy downtown locker (Standard) to the remote vault (Archive) — and unlike rivals, even the deep vault hands your box back almost instantly rather than making you wait hours.

Key Concepts

    gs://my-bucket/images/2026/06/photo.jpg
                   ^^^^^^^^^^^^^^^^^^^^^^^ one object name

The distinguishing trait: one API across every class.

    Standard   hot data
    Nearline   ~monthly access, 30-day minimum
    Coldline   ~quarterly, 90-day minimum
    Archive    yearly, 365-day minimum

Archive is still milliseconds to first byte. There is no rehydration step and no waiting — a real difference from competitors, where archival tiers are offline and take hours. The cost of cold classes is in retrieval fees and minimum durations, not in access time.

Early deletion is charged. Delete a Nearline object after a week and you still pay 30 days.

Location type is chosen at creation and cannot change.

    region        one region, cheapest, lowest latency locally
    dual-region   two specific regions, synchronous-ish
    multi-region  a continent, highest availability

Autoclass is the easy win. Enable it and objects move between classes based on actual access, with no lifecycle rules to maintain and no early-deletion surprises.

Lifecycle rules otherwise do it by age or condition — Standard to Nearline at 30 days, Coldline at 90, delete at a retention limit.

Access control: prefer uniform.

    uniform bucket-level access   IAM only. Recommended.
    fine-grained                  IAM plus per-object ACLs. Legacy,
                                  and how objects get made public
                                  by accident.

Signed URLs give time-limited access without credentials, which is how browser uploads and private downloads work.

Public access prevention can be enforced at the organisation level so a world-readable bucket is impossible rather than merely discouraged.

It is the hub of the data platform — the staging area for BigQuery loads, Dataflow sources and sinks, and the destination for log exports.

What the interviewer is probing.1. "How does Archive differ from other clouds' archival tiers?" Probing: the access model. Stalls: "It is cheaper and slower." Moves up: first-byte latency is still milliseconds — there is no rehydration step, so the cost is in retrieval fees and minimum durations rather than waiting hours.

2. "What is Autoclass and why use it?" Probing: the easy saving. Stalls: "A storage class." Moves up: it moves objects between classes based on actual access with no lifecycle rules to maintain and no early-deletion charges.

3. "Uniform or fine-grained access control?" Probing: the recommendation and why. Stalls: "Fine-grained is more flexible." Moves up: uniform bucket-level access — IAM only; per-object ACLs are how objects end up public by accident.

4. "Can you change a bucket location type later?" Probing: the permanent decision. Stalls: "Yes, in settings." Moves up: no — region, dual-region or multi-region is fixed at creation, so it needs deciding up front.