While VMware highly recommends the deployment of all vCenter Server components into a single virtual machine (excluding the vCenter Server database), large enterprise customers running multiple vCenter Server instances within a single physical location can simplify the vCenter Single Sign-On architecture and management by reducing the footprint and required resources and specify a dedicated vCenter Single Sign-On environment for all local resources in each physical location.
For vSphere 5.5 the VMware recommendation is to centralize vCenter Single Sign-On when you have 8 or more vCenter Server instances in a given location (this is a soft recommendation).
Centralized vCenter Single Sign-On Architecture
There can be increased risk when centralizing a vCenter Single Sign-On server (to why it is not recommended for smaller environments) due to the increased number of components affected if the vCenter Single-Sign-On server was to become unavailable, in short all vCenter Server components of all vCenter Servers registered will incur authentication loss (when compared to just the single vCenter Server instance when installed locally) and so availability of the vCenter Single Sign-On centralized server(s) is highly recommended.
Centralized Single Sign-On High Availability options
The absence of vCenter Single Sign-On server greatly impacts the management, accessibility and operations within a vSphere environment. The type of availability required is based on your Recovery Time Objective (RTO), and VMware solutions can offer various levels of protection.
VMware vSphere Data Protection
VMware vSphere Data Protection™ provides a disk-level backup and restore capability utilizing storage-based snapshots. With the release of vSphere Data Protection 5.5, VMware now provides the option of host-level restore with vSphere Data Protection and users can backup the vCenter Single Sign-On server virtual machine(s) using vSphere Data Protection and can employ the capability later to restore as necessary to a specified vSphere host.
When deploying a centralized vCenter Single Sign-On server to a vSphere virtual machine environment, users can deploy vSphere HA to enable recovery of the vCenter Single Sign-On server virtual machines. vSphere HA monitors virtual machines via heartbeats from the VMware Tools™ package, and it can initiate a reboot of the virtual machine when the heartbeat is no longer received or when the vSphere host has failed.
vCenter Server Heartbeat
VMware vCenter Server Heartbeat™ provides a richer availability model for the monitoring and redundancy of vCenter Server and its components. vCenter Server Heartbeat places a centralized vCenter Single Sign-On server into an active/passive architecture and will monitor the application as well as provide an up to date passive node for recovery during a vSphere host, virtual machine or application failure.
Network Load Balancer (NLB)
A VMware or third party Network Load Balancer (NLB) can be configured to allow SSL Pass Thru communications to a number of local vCenter Single Sign-On server instances and provide a distributed and redundant vCenter Single Sign-On solution. Although VMware provides network load balancer capability in some of it’s optional products, such as vCloud Networking and Security, , there are also third party solutions available in the market. VMware does not provide support for the third party NLB solutions.
This information is from the upcoming ‘VMware® vCenter Server™ 5.5 Deploying a centralized vCenter Single Sign-On server with a Network Load Balancer (NLB) Technical Reference Guide that will highlight configuration examples of Network Load Balancers (NLB) with vCenter Single Sign-On.