Cloud Pod Architecture is not a load-balancing solution, and therefore if multiple sites are configured in a federated Cloud Pod, a global site load balancer must be implemented. The global load balancer must be configured with a single namespace for the tenant (for instance, tenantpod1234.myvmware.com) which will direct the user connection to one of the Access Point appliances in either site.

Local load balancing must be in place for each service provider data center, in order place connections on an available Access Point appliance for a given site. Depending on the load balancer used, a number of algorithms can be configured, but in most cases “least connections” is preferred.

It is also recommended that the load-balancer monitors the health of each Access Point appliance by sending an HTTP GET /favicon.ico request. This will check for the presence of /favicon.ico served by the Access Point web server. If a "200 OK" response is received, the load balancer will consider the Access Point node to be heathy. If there is no response, then the load balancer will consider it offline and not send further connections to the offline Access Point appliance.

Some of the most common methods for providing a load balanced external connection is to use a single virtual IP (VIP), provide multiple VIPs, or use the single VIP with port forwarding.

For more information on load balancing please refer to “Load Balancing with VMware Access Point” (see Section 10, References).