Architecting a vCloud Director Solution : Resource Groups : 6.4 Storage : 6.4.1 Storage Tiering
   
6.4.1 Storage Tiering
To handle the elasticity and scalability required for a vCloud implementation, a modular tiered storage approach is recommended. This involves designing for future growth while optimizing for performance. Storage differentiation for each service tier is determined, and then pools of tiered storage are presented to vCloud Director.
The following are storage tiering design considerations:
VM storage policies must be enabled on every resource group vSphere cluster.
User-defined storage tags (for example, bronze, silver, gold) are created and assigned to vSphere datastores. Each vSphere VM storage policy is mapped to corresponding tags with a set of rules. Alternatively, capability-based storage policies can be created for storage resources that support them. Upon synchronization, the storage policies can be assigned to provider VDCs in vCloud Director. In general, do not the universal storage profile * (any). (If it is used, disable the local datastore.)
vCloud Director references datastores by the vSphere storage policy names. Do not rename them afterwards in vSphere.
A single vCloud Director VM can span multiple datastores. For example, VM disk 1 can be on datastore A and VM disk 2 on datastore B, where datastores A and B belong to the same or a different storage policy. When fast provisioning is enabled at the Org VDC level, the VM cannot span multiple datastores.
A single disk can be attached to only one VM at a time. For shared storage between VMs, in-guest IP storage can be used (NFS, iSCSI).
Independent disks (available only through the vCloud Director API) can be easily moved between VMs.
Storage policies can be used for placement of VMs to a subset of ESXi hosts due to licensing restrictions (Windows as opposed to Linux hosts). Storage policy of a particular type (for example Linux) is assigned to a datastore that is attached only to hosts licensed for Linux workloads.