One of the most important and well received features of the new vSphere 5 release, Storage DRS, automates I/O and storage capacity management, while making VM provisioning much easier. Though admins need to continue to manage storage tiers and provision enough performance capacity in each datastore cluster, this new feature is a significant step forward. Some of the features that Storage DRS brings to the table:
The datastore cluster is a new object comprised of like datastores that share many of the same characteristics (such as RAID, VMFS, and NFS). For example, datastores with RAID5 and RAID10 would not exist in the same datastore cluster, and replicated and nonreplicated datastores would not be mixed. Datastores in a datastore cluster do not have to be on the same array or from the same vendor, but do need to share the same configuration characteristics (helpfully, NFS is supported). They can also have different sizes and I/O capacities, but should be of the same service level or tier. One or more host clusters can connect to a datastore cluster, and a single host cluster can connect to multiple datastore clusters.
It’s now much easier to accommodate future growth and incrementally expand storage by adding datastores to an existing cluster, rather than having to buy a large storage array from the start.
Once you have a datastore cluster configured and you need to provision or migrate a VM, you begin to see the benefits of initial placement. During these processes you’ll choose the datastore cluster based on the service level or tier required. The VM and VMDKs are intelligently placed on a datastore in the datastore cluster based on I/O load and space utilization.
Datastore maintenance mode is basically the same as host maintenance mode: In daatstore maintenance mode, the VMs and VMDKs from the selected datastore move to another part of the datastore cluster.
Automated I/O load and space balancing is where Storage DRS gets more complicated. User-defined thresholds for I/O response and space utilization create an algorithm that determines migration recommendations. If the threshold is exceeded, the predefined action is triggered accordingly—a migration or recommendation, automatically or manually.
VMDK affinity is on by default, which places all VMDKs for a particular VM on the same datastore. What’s great is the new capability to separate a VM’s VMDK files using a VMDK anti-affinity rule. This is great for separating logs and data for database VMs. VM anti-affinity can now keep VMs and associated VMDKs separate from other VMs and VMDKs. This comes in handy for disaster recovery.
Storage DRS simplifies VM guest provisioning, makes it easier to incrementally add storage, simplifies datastore I/O balancing, and provides much more flexibility for VMDK placement. It still requires detailed knowledge of the storage infrastructure for datastore management, but is a significant step forward in making VM provisioning easier.
Unique control with VM-level actions for infrastructure functions including snapshots, replication and QoS make protection and performance certain in production, and accelerate test and development cycles.