Production deployments of vCloud Availability have specific sizing and component configurations.

A production deployment uses multiple components to support many protected virtual machines, and to provide a fault-tolerance within the DRaaS environments.

Production Architecture

Production deployments must meet certain requirements.

  • At each tenant site, there is one or more single-tenant environments to be protected.

  • In the service provider disaster recovery site, one or more vCloud Director is configured with a specific number of components designed to handle the required number of VMs from each tenant.

  • A single vCloud Director environment in a data center hosts up to 1000 individual tenants.

Using this information as a base, you must install and configure a new vCloud Availability service instance for every 100 tenants that use the service.

In a single vCloud Director deployment, there is a limit to the number of VMs that can be replicated as part of the DR solution. The exact combination depends on the number of VMs that must be supported combined with the system limits for each component.

Component Sizing

Individual components have a minimum installation count required for a base installation.

Table 1. Relative Component Sizing

Component

Related Component

vCloud Director

1 Management vCenter Server Instance

1 Resource vCenter Server Instance

vSphere Replication Manager

4 vSphere Replication Server

Component Limits

Individual components have limits for the maximum number of supported services, instances, or connections required.

Table 2. Component Counts and Limits per Instance

Component

Limit

Cloud Proxy

2000 Connections (vRealize Automation control connections and replicated VM payload)

vSphere Replication Server

250 Active Replications

Tenants

1000 per vCloud Director

vCloud Director

15 vCenter Server Instances

vCloud Availability Portal

800 Concurrent Sessions

Sample Deployment Scaling

Using the information on sizing and configurations, for a single instance, supporting up to the maximum of 100 tenants.

Table 3. Component Counts for Production Deployments

Protected VMs

500

1000

2000

3000

5000

10000

vSphere Replication Server

12

24

48

60

120

240

vSphere Replication Cloud Service

2

2

3

3

3

3

Cloud Proxy

2

2

2

2

3

5

Minimum High Availability Component Counts

To sustain a highly available vCloud Availability solution, consider the following counts per component.

Table 4. Component Counts for Highly Available vCloud Availability Solution

Component

Number of Instances

vCloud Director

2

vCenter Server

1 Management vCenter Server Instance with 2 ESXi hosts.

1 Resource vCenter Server Instance with 2 ESXi hosts.

External Platform Services Controller (PSC)

1

vSphere Replication Manager

1

vSphere Replication Cloud Service host

2

vSphere Replication Server

4

Cloud Proxy

2

vCloud Availability Portal

2

vCloud Availability Administration Portal

1

Cassandra Host

3

RabbitMQ Host

2

Network Configuration Best Practice

Within a production deployment, the underlying network architecture is important. The following describes a best practice network configuration.

  • Each underlying physical ESXi installation is configured with a VMXNET3 high-speed network adapter, connected to two separate 10Gbe switches using NIC teaming.

  • The two switches are connected to each other using two 40gbe QSFP cables.

  • The switches are configured to present a VLAN that is configured as a port group on the ESXi hosts.

  • The virtual machines that make up the environment are all configured in this flat broadcast domain.