Assuming you're talking about Datastore Cluster, the best practice is to have all volumes on a shared storage, this way the vSphere Storage DRS will be able to balance virtual machines between datastores automatically. If you use local disks, the virtual machine files availability will be affected in case of a host failure.

If you plan to have High Availability is not a good idea to use local disks, since that disks are not shared between hosts, and if a single host fails, all virtual machines stored on local disk of that particular host will become unavailable.

The advantage of a datastore cluster is to manage and manage them as a single entity and of course provide load balancing based on space and/or I/O latency and create anti-affinity rules. You can read more about benefits here: An introduction to Storage DRS