Storage Distributed Resource Scheduler (SDRS) was introduced in vSphere 5.0 to efficiently manage a pool of datastores as a single logical datastore (a datastore cluster). VM optimization and distribution were based on two metrics—space and I/O.

SDRS fully supports VMFS and NFS datastores. However, it does not allow adding NFS datastores and VMFS datastores into the same datastore cluster. Starting with vSphere 6.0, SDRS is now aware of the storage capabilities available through VASA 2.0 and can use storage policies (see later in this chapter). It will only move or place VMs on a datastore within the datastore cluster that can satisfy specific VM's storage policies, based on several features, including the following:

• Deduplication: SDRS will be aware of deduplication domains, and when datastores belong to the same domain, moving a VM will have little to no effect on capacity.
• Storage tier: SDRS will not move VMs while the storage auto-tier is promoting or demoting blocks to a lower or higher tier.
• Thin provisioning: SDRS can recognize if more thin-provisioned datastores have a common backing pool to avoid migrating VMs.
• Storage replica (SR): SDRS will recognize replica VMs and avoid resource constraints between storage vMotion and replication.

Always refer to the storage vendor's guides to verify if SDRS is supported and with which settings (usually auto-tiering storage is incompatible with SDRS I/O balance).

Also note that SDRS cannot replace the tiering mechanism implemented in some storage (especially hybrid storage); just because a single VMDK can only stay in one single datastore and not split across multiple datastores, there is no way for SDRS to move hot blocks to a fast datastore and cold blocks to a slow datastore.

Datastore cluster and SDRS management remain the same as in vSphere 5, with similar considerations, such as keeping disks with the same capabilities in the same datastore cluster. However, with vSphere 6.0 you can now mix datastores with different capabilities in the same datastore cluster, using the right storage policies at the VM level.