Owner's Manual

HA and SDN
The HA protocol provides redundancy and hitless failover for the AFC during upgrades or failure. Dell
Networking recommends using at least two controller servers configured for HA (one active, one
standby) to enable this feature. To enable this feature on a single controller server, you can temporarily
assign another server as a standby server, then initialize failover by disabling the current active server.
When the upgrade completes, you can remove or reassign the server.
The active and standby controllers remain in sync using periodic checkpoints. High-level information,
such as the last calculated topology and a list of end hosts, is included in the checkpoint data. The active
controller hosts all network applications, such as the database server. The active controller can run as a
VM, sharing the physical server with other components such as OpenStack. The standby controller can
join or leave the group as needed without disrupting traffic. To ensure the standby controller has enough
checkpoint data to become an active controller, it initializes a reconciliation process before becoming
active. After this verification process completes, the controller’s role changes to active without resetting
any of the forwarding planes.
Deployment Model
The following example depicts a typical deployment with HA. The key components of this deployment
are:
• AFC controllers are running in Active/Standby mode.
• The GUI uses AFC virtual IP (VIP) for REST communication and is unaware of the active and standby
AFC controllers.
• The switches have two OF connections (one active, one standby).
• The switches send control and data packets over the OF connection to the active controller.
22
Supported Technologies and Protocols