VMware Engine private clouds
A Google Cloud VMware Engine private cloud is an isolated VMware stack that consists of the following VMware components:
- ESXi hosts
- vCenter Server
Private clouds help you address a variety of common needs for network infrastructure:
- Growth. Add nodes with no new hardware investment when you reach a hardware refresh point for your existing infrastructure.
- Fast expansion. Create additional capacity immediately when temporary or unplanned capacity needs arise.
- Increased protection. Get automatic redundancy and availability protection when using a private cloud of three or more nodes.
- Long-term infrastructure needs. Retire data centers and migrate to a cloud-based solution while remaining compatible with your enterprise operations. This is especially useful if your data centers are at capacity or you want to restructure to lower costs.
Private cloud environment
You manage your private clouds through the VMware Engine portal. Each private cloud has its own vCenter Server in its own management domain, and all nodes in a given private cloud reside in the same region.
The VMware stack runs on dedicated, isolated bare metal hardware nodes in Google Cloud locations. You use the stack through native VMware tools, including vCenter Server and NSX Manager.
Private clouds are also designed to eliminate single points of failure:
- Clusters of ESXi hosts are configured with vSphere High Availability (HA) and sized to have at least one spare node for resilience. vSphere HA protects against node and network failures.
- vSAN provides redundant primary storage. vSAN requires at least three nodes in a private cloud to provide protection against a single failure. You can configure vSAN to provide higher resilience for larger clusters.
You can connect the private cloud to your on-premises environment using the following connections:
Single-node private clouds
For pilot testing and proofs of concept with VMware Engine, you can create a private cloud that contains only a single node and cluster in any region where VMware Engine is available. All VMware Engine features are available in a single-node private cloud, but there are specific limitations on VMware stack features due to cluster size.
The following are common use cases for a single-node private cloud:
- Proof of concept: evaluating VMware Engine and its capabilities
- Disaster recovery testing: deploying your application from recent backups to periodically validate disaster recovery preparedness
- Application upgrade testing: test and validate application component upgrades before upgrading your application in production
VMware Engine deletes single-node private clouds after 60 days, along with any associated workload VMs and data. However, you can prevent this deletion and receive production-ready SLA coverage by expanding to at least 3 nodes within 60 days. The expansion process won't disturb your VMs or access to vCenter, and it initializes vSAN data replication after the nodes are successfully added to the cluster.
For single-node private clouds, the default vSAN storage policy uses a Failures to Tolerate (FTT) value of FTT=0. When you expand a single-node private cloud, VMware Engine changes the default vSAN storage policy. The default vSAN storage policy changes to use FTT=1 for 3–4 node private clouds and to use FTT=2 for private clouds with at least 5 nodes.
Custom core counts
Some licensing agreements, such as those for Windows Server and SQL Server, charge you based on the number of CPU cores on the underlying physical node or in the cluster. Whenever you create a new cluster, you can reduce the number of available cores for each node in the cluster to meet the application license requirements. VMware Engine also creates any new nodes added to that cluster with the same number of cores per node, including when replacing a failed node.
Custom core counts are available for both the initial cluster as well as for any other clusters created in a private cloud. Reducing the number of available cores doesn't affect node pricing.
Each private cloud has resource limits for its nodes and clusters. Refer to VMware in a private cloud for a list of these limits.
Single-node private cloud limitations
For single-node private clouds, the following limitations apply to the VMware stack:
- Features or operations that require more than 1 node won't work. For example, you won't be able to use vSphere Distributed Resource Scheduler (DRS) or High Availability (HA).
- The default vSAN storage policy uses FTT=0, so node failure results in data loss.
Additionally, the following VMware Engine limitations apply:
- You cannot add a single-node cluster to an existing private cloud.
- An existing private cloud cannot be converted to a single-node private cloud.
- Node adjustments by autoscale policies aren't supported with a single-node private cloud.
- VMware stack upgrades involve a downtime for your single-node private cloud.
- VMware Engine doesn't encrypt data on vSAN on a single-node private cloud by default.
- A private cloud must contain at least 3 nodes and complete vSAN data replication to be eligible for coverage based on the SLA.
- You cannot adjust the number of cores per node with a single-node private cloud.
Custom core count limitations
The following limitations apply to a cluster that has a custom core count:
- All nodes added to the cluster after initial creation also use the custom core count.
- The number of cores per node can't be changed after cluster creation. To change the number of cores per nodes in the cluster, you must delete the cluster and create a new one.
- The number of cores per node must be a multiple of 4 (like 4, 8, 12, or 16).
- Custom core counts aren't available for single-node private clouds.
- Learn about VLANs and subnets.