Compute and Storage Services
Compute choices
Compute Instances give you a full OS and root access, and are the default answer for self-managed software. LKE provides a managed Kubernetes control plane while you pay only for the worker nodes and their attached storage. App Platform builds and deploys applications from a Git repository and bills by plan tier, which suits teams that do not want to manage servers. Images let you capture an existing disk as a reusable template, and Marketplace/StackScripts automate first-boot configuration.
Storage choices
Block Storage volumes are network-attached disks that must live in the same data center as the instance that uses them. You attach a volume to one instance at a time, must create a partition and filesystem on it before use, and can grow it but never shrink it. Unattached volumes are still billed.
Object Storage is S3-compatible, organizes data into buckets, and is accessed with access keys and HTTP APIs rather than mounted as a disk. It suits static assets, logs, artifacts, and backup targets.
Backups are scheduled daily and weekly snapshots of an instance's disks, billed as a percentage surcharge on the instance. They are not the same thing as an image you create manually, and they are not free.
Traps to watch
- A volume created in the wrong region or data center simply cannot attach to your instance.
- Object Storage cannot be mounted as a filesystem; questions that imply otherwise are traps.
- Deleting an instance destroys its local disks and any backup history attached to it.
- Resizing a volume downward is unsupported; plan capacity growth deliberately.
Rule of thumb
Put block storage under databases, mail spools, and application data that needs a POSIX filesystem; put Object Storage under media, build artifacts, and off-instance backups. Never put a live database on Object Storage.