A reservation type VDC is used for a dedicated service offering where the tenant obtains its own compute resources (vSphere cluster). The tenant can set reservations, limits, and shares at the VM level and manage resource oversubscription.

The maximum size of the VDC is the useable cluster capacity (after taking out HA N+1 and hypervisor overheads) lowered by necessary resource for edge gateway deployment.

The edge gateways might be deployed into provider VDC secondary resource pools, which is useful in NSX Edge cluster scenarios (see Section 6.3.2.3, Design Option 2b – Combined Edge/Compute Cluster with Non-Elastic VDC). In this scenario, the cluster for edge gateways can be shared among many different provider VDCs by dividing it into manually created resource pools.

In the following figure, the primary resource pool of each provider VDC is a dedicated cluster, which can be as small as two hosts. The NSX Edge cluster contains multiple small resource pools, which are added to each Reservation Org VDC as secondary resource pools.

This design allows use of highly available NSX Edge instances (active/passive) when a tenant wants to buy only two host clusters and pay software licensing (for example, database software) for one host with the second used only for failover. The vSphere HA admission control is then set to the dedicated failover host.

 

VMware does not recommend using a reservation type VDC in a shared cluster. The tenant can create an arbitrarily large VM because its physical compute resources are limited by the Org VDC resource pool and not by allocated VDC resources. (For example, inside an Org VDC allocated with 100 GB RAM, it is possible to power on a VM with 200 GB RAM.) If the VM memory is larger than the resource pool limit, the ESXi host running the VM will most likely result in swapping the memory it cannot physically provide. The swapping can create extensive load on the host and its storage, and possibly have a negative impact on other tenant workloads.