Virtualized Cluster

Private Cloud Director enables you to manage and operate multiple virtualized clusters from a single PCD region.

Use Cases

  1. Multi-Tenant Isolation - You can assign different tenants to separate clusters to enforce resource boundaries and fault domain separation. The isolation prevents noisy neighbor issues and enhances security between tenant environments.

  2. Licensing Requirements - Some specialized software, such as Oracle database, may require isolating hosts with the software license enabled. Creating a separate cluster will allow you to do this.

  3. Hardware Specialization - You can group hosts with similar capabilities (such as GPU-enabled or high-memory machines) into dedicated clusters to optimize performance and resource utilization for specialized workloads.

Understanding a Virtualized Cluster

A virtualized cluster is a grouping of interconnected physical servers or hosts, with certain cluster level features operating at the group level. Resources such as CPU, Memory, and GPU across all hosts in a virtualized cluster are presented as a single pool, so that when you create a new virtual machine in the virtualized cluster, the placement of the new virtual machine may be on any of the underlying physical servers, based on capacity and other provisioning constraints.

All clusters within a region operate under a single cluster blueprint, ensuring consistency in base configurations while allowing for cluster-specific customization.

Each virtualized cluster provides two key features that enable you to run production workloads.

Virtual Machine High Availability (VM HA)

VM HA provides automatic fault tolerance for your workloads. When enabled:

  • The system continuously monitors host health across the cluster.

  • If a host failure is detected, affected VMs are automatically recovered on healthy hosts.

  • Minimize downtime without requiring manual intervention.

  • Business continuity is maintained even during infrastructure issues.

Read more here about Virtual Machine High Availability (VM HA).

Dynamic Resource Rebalancing (DRR)

DRR works as a continuous optimization engine that:

  • Monitors allocated capacity and real-time utilization metrics (CPU and memory) across all hosts in the cluster.

  • Analyzes resource distribution patterns to identify imbalances

  • Intelligently migrates VMs across hosts in the cluster to optimize resource utilization.

  • Prevents hotspots and resource contention before they impact performance

This proactive approach ensures that your clusters maintain optimal performance even as workload patterns change over time.

Read more here about Dynamic Resource Rebalancing (DRR).

GPU

Enabling GPU support allows you to group all hosts capable of GPU passthrough or vGPU-backed hypervisors into a single GPU-aware virtual cluster.

Before configuring GPU mode on any GPU hosts, ensure the following points are met.

  • Supported GPU Modes: Passthrough or vGPU

  • Use Passthrough to assign full physical GPUs to virtual machines.

  • Use virtual GPUs (vGPUs) to virtually slice existing physical GPUs and share them across virtual machines.

  • Non-GPU hosts can not be added to a GPU-enabled cluster.

This setting ensures that all GPUs in the cluster use the same GPU mode (Passthrough or vGPU) for uninterrupted GPU operations, such as live migration between GPU hardware, resizing, and scaling GPUs.

Read more here about GPU Support in Private Cloud Director.

CPU Mode and Model

Defines how the hypervisor exposes CPU features to virtual machines. You can select one of the following CPU modes during cluster creation in the Private Cloud Director UI.

  • Default: Automatically selects the latest supported CPU model on the host.

  • Host Model: Matches the physical CPU closely, balancing performance and migration flexibility.

  • Host Passthrough: Exposes the physical CPU directly for maximum performance (migration only possible between identical hosts).

  • Custom: Specify a standardized CPU model to ensure consistent virtualized CPUs across hosts and enable live migration between different hardware.

Create a Virtualized Cluster

To create a new virtualized cluster:

  1. Navigate to Infrastructure > Clusters > Add Cluster

  2. Provide a name for the cluster.

  3. Choose desired settings for:

    1. VMHA (Virtual Machine High Availability)

    2. DRR (Dynamic Resource Rebalancing)

    3. GPU (Passthrough or vGPU)

PreviousSetting Up a Console ProxyNextDynamic Resource Rebalancing (DRR)

Last updated

Was this helpful?