One type of VxRail cluster is where the VxRail nodes provide all the physical compute and storage resources to support application workload, and the primary storage resource is vSphere vSAN. For this cluster type, the slots in the nodes are filled with disk drives that meet the performance and capacity requirements for the application workload and then formed into a local vSAN datastore during the cluster initialization process.
Figure 15. Local vSAN datastore deployed on a VxRail cluster to be the primary storage resource
The cluster initialization process performs an inventory of the disk drives on the nodes and uses that discovery process to form a vSAN datastore. The high-endurance SSD drives discovered on each node serve as a cache for virtual machine I/O operations. The high-capacity drives discovered are the primary permanent storage resource for the virtual machines. The vSAN formation process partners the cache drives with one or more capacity drives to disk groups on the nodes, with the resulting vSAN datastore consisting of the collection of disk groups.
This cluster type is designed to handle most customer application workloads and covers the most common use cases.
- With a local vSAN datastore, operational flexibility can be realized to address scalability and high availability requirements.
- This cluster type is simple to deploy and operate. The initialization process performs all the work to pool all the node resources for ease of consumption
- This cluster type is not dependent on external resources for storage, so all resources can be administered under a single point of management.
- Additional compute and storage resources can be expanded easily to the cluster through automated node and disk drive addition.
- External storage resources can be configured on the cluster as secondary storage capacity.
However, this cluster type may not be a good fit for certain use cases:
- This cluster type may be cost-prohibitive for smaller business requirements. Light workloads, such as those in a remote office, may be a better fit for a two-node cluster or a satellite node.
- This cluster type is in a single site. It offers support for continuous operations from a failure of components within the cluster but does not offer zero-downtime protection from a failure of a single data center or site. A stretched VxRail cluster is a better solution for very high availability requirements.
- Expansion of a cluster through node addition can potentially lead to stranded assets, where excess compute and storage resources cannot be shared outside of the cluster. For workloads which require a more precise balance of compute and storage resources, a dynamic cluster may be a better fit.