Part 8 of the 10-part VCF series. This technical reference manual outlines the structural components, deployment workflows, and core design principles governing Workload Domains within a VMware Cloud Foundation (VCF) 9.x architecture.
Workload Domain Core Component Stack
A Workload Domain represents a policy-driven pool of compute, storage, and network capacity managed by a dedicated control plane. Unlike standalone vSphere clusters sharing a single management footprint, each VCF workload domain receives:
- A dedicated, isolated vCenter Server appliance.
- One or more vSphere clusters consisting of certified ESXi hosts.
- Localized NSX networking configurations (either mapped to a dedicated local three-node NSX Manager cluster or mapped to a shared NSX management instance to minimize resource usage).
This strict control-plane separation isolates administrative faults, configurations, and lifecycle upgrade windows, restricting the blast radius of any infrastructure event exclusively to that domain.
Orchestrating Workload Domains via SDDC Manager
Workload domain automation is managed entirely through SDDC Manager. The deployment workflow requires specific prerequisite configurations:
- Host Commissioning: Validating physical hardware and assigning IP ranges to the host pool via Network Pools.
- Resource Definition: Inputting the unique fully qualified domain name (FQDN) and resource parameters for the new tenant vCenter Server.
- Principal Storage Assignment: Designating the storage profile, selecting between native vSAN ESA/OSA, or connecting pre-provisioned external Fibre Channel, iSCSI, or NFS arrays.
- Network Fabric Overlay: Mapping Distributed Virtual Switches (VDS), uplink profiles, and specifying the corresponding NSX Transport Zones.
Once initialized, SDDC Manager fully automates the vCenter deployment, assigns the ESXi hosts to new clusters, handles storage mounting, hooks up the transport nodes to the NSX overlay, and applies the designated global licensing.
Importing Existing Greenfield and Brownfield Clusters
VCF 9.x includes native automation paths to ingest existing vSphere infrastructure into the managed cloud topology via SDDC Manager.
- Minimum Baseline: Source environments must run a minimum of vSphere 8.0 Update 3 or higher to qualify for the brownfield migration path.
- Automation Hook: When importing a standalone vCenter instance, the SDDC Manager workflow ingests the inventory, establishes compliance hooks, and automatically provisions a new three-node NSX Manager cluster if an existing NSX environment is not already active on the target virtual distributed switches.
Multi-Cluster Topologies and Workload Segmentation
A single Workload Domain can scale horizontally to support multiple distinct vSphere clusters. This multi-cluster approach allows administrators to segment diverse compute requirements while maintaining a single vCenter management boundary:
- Standard Compute: Classic x86 nodes running standard virtual machines.
- Accelerated Compute: Specialized clusters running hardware accelerators (e.g., NVIDIA GPUs) dedicated to Private AI Services.
Both clusters share the unified NSX infrastructure and vCenter boundary of the parent domain but operate on completely independent hardware host types and vSphere Distributed Resource Scheduler (DRS) rules.
Architectural Sizing and Scale Limits
The choice to deploy a new Workload Domain versus expanding an existing domain via an additional cluster must be driven strictly by technical isolation requirements:
- Separate Domains: Required if workloads necessitate independent upgrade windows, separate security zones, or completely isolated administrative credentials.
- Multi-Cluster Extensions: Preferred if workloads can safely share a single vCenter instance and upgrade schedule, mitigating the resource overhead of deploying superfluous vCenter and NSX appliances.
VCF 9.x scales to support up to 5,000 ESXi hosts per single SDDC Manager instance, with optimized multi-threading capable of executing concurrent, parallel host lifecycle management tasks across multiple clusters inside the fleet.