Google Cloud Platform

GCP Cloud Storage

Binary Large Object Storage - Fully managed, scalable.

Can be used to serve website content, archive data or content for download.
Each object has a URL.
Stored in buckets.
Immutable.
Can create ACLs (access control lists) - A scope (who) & permission (what).
New overwrites old but can keep previous versions (object versioning) - With rules on how many.
Lifecycle management policies can delete old versions of objects or simply old object.

Multi-regional - 99.95% - Frequent access. Stored in 2 locations > 160km apart. (6 copies of data round the world).
Regional - 99.90% - Frequent access. (3 copies of data in a Region).
Nearline - 99.00% - Access less than once a month. A cost to access.
Coldline - 99.00% - Access less than once a year. A cost to access. Quick (AWS Glacial is 4-6 hours). Costs per GB per Month stored. For Nearline, Coldline also a cost to access.

Projects own Buckets (the basic containers for all data).
Buckets have a default storage class.
Buckets cannot be nested.
There are limits to Bucket creation and deletion so favor Object or Bucket operations.
Have a globally unique name name, a geographical location and a default storage class (can override class though).
Name and location are fixed.
The existence of a bucket name can be globally tested so avoid usernames, email, PII etc.
Max 63 characters unless include . (dots) then can be 222 but max 63 between dots.
Name format: [a-z][0-9][-_.]
Bucket Labels: name:value pairs.
IAM can be applied to Project or Folder.

Names can contain any UTF-8 char.
(obv) avoid characters like / in names.
Every object has a generation number. Incremented each time object overwritten.
If object versioning enabled can refer to prev versions.
Objects have two version numbers, one for content the other for metadata.
Objects are opaque to GCP and immutable.
Cloud storage does not support a hierarchy but tools like GCP console can operate as if in a hierarchy.