Modelling Object Storage

[Alan Prochaska]

For us, for traditional storage we model arrays as storage servers, allocations as storage volumes which are accessed as file systems and related to the servers on which they're mounted.  Storage volumes are related to the storage server from which they're allocated and to the server on which the file system is mounted.  The storage servers are managed by a technical service

Object storage presents some differences.  Our object storage comes from a SaaS service, so the servers or arrays providing the storage are hidden.  We receive storage volumes from the technical service.  But object storage volumes (aka buckets) can be used directly; they need not be mounted on a server.  Thus, we'll relate them directly to an application service, not to a server.

This suggests an object storage volume should have different relationships than a traditional storage volume.  It should have a technical service and an application service as parents.  We're thinking of subclassing Storage Volume (ie. extend table cmdb_ci_storage_volume) to Object Storage Volume (u_cmdb_ci_object_storage_volume) so we can manage them separately.  Among other things, Object Storage Volume will have different orphan rules.

Thoughts, anyone?

Has anyone else modelled object storage in a way that works?

Thanks for any thoughts and feedback.

Alan Prochaska