A reservation is a share of available memory, CPU and storage one compute resource that is reserved for use by a particular fabric group. Each reservation is for one fabric group only but the relationship is many-to-many. A fabric group might have multiple reservations on one compute resource, or reservations on multiple compute resources, or both. A reservation must include a vSphere resource pool.

Shared Edge and Compute Clusters and Resource Pools

While reservations provide a method to allocate a portion of the cluster memory or storage in vRealize Automation, reservations do not control how CPU and memory are allocated during periods of contention on the underlying vSphere compute resources. Use vSphere resource pools to control the allocation of CPU and memory during time of resource contention on the underlying host. To fully use the mechanism of resource pools for provisioning of workloads, all VMs must be deployed on one of the following resource pools.

Table 1. Resource pool details

Resource Pool

Object Types

sfo01-w01rp-sddc-edge

NSX Edge components at the data center level. Place user workload in other resource pools.

sfo01-w01rp-user-edge

Statically or dynamically deployed NSX components such as NSX Edge gateways or load balancers which serve specific customer workloads

sfo01-w01rp-user-vm

Statically or dynamically deployed virtual machines such as Windows instances, Linux instances, databases, etc., which contain specific customer workloads

Table 2. Design Decisions on Reservations in vRealize Automation

Decision ID

Design Decision

Design Justification

Design Implication

SDDC-CMP-036

Create at least one vRealize Automation reservation for each business group at each region.

In this design, each resource cluster has two reservations: one for production and one for development, so that you can provision both production and development workloads on the cluster.

Because production and development share compute resources, the development business group must be limited to a fixed amount of resources.

SDDC-CMP-037

Create at least one vRealize Automation reservation for edge resources in each region.

NSX can create edge services gateways on demand and place them on the shared edge and compute cluster.

The workload reservation must define the edge reservation in the network settings.

SDDC-CMP-038

Configure all vRealize Automation workloads to use the sfo01-w01rp-user-vm resource pool.

You introduce control over the resources allocated to the tenant application workloads. As a result, using another resource pool you can dedicate more compute resources to the NSX networking components that provide networking to the workloads.

Workloads provisioned at the root resource pool level receive more resources than those in child resource pools. In contention situations, virtual machines might receive insufficient resources.

Cloud administrators must ensure that all workload reservations are configured with the right resource pool. You can have a single resource pool for both production and development workloads, or two resource pools, one dedicated for the Development Business Group and one dedicated for the Production Business Group.

SDDC-CMP-039

Configure vRealize Automation reservations for dynamically provisioned NSX Edge components (routed gateway) to use the sfo01-w01rp-user-edge resource pool.

You dedicate compute resources to NSX networking components. You must assign a vSphere resource pool to end-user deployed NSX Edge components.

Workloads provisioned at the root resource pool level receives more resources than child resource pools, which might starve those virtual machines in contention situations.

Cloud administrators must verify that all workload reservations are configured with the right resource pool.

SDDC-CMP-040

All vSphere resource pools for edge or compute workloads must be created at the root level.

Do not nest resource pools.

Nesting of resource pools can create administratively complex resource calculations that might result in unintended under- or over-allocation of resources during contention situations.

None.