Design the wireless network for high availability
Domain 3: Design the wireless network for high availability
This domain tests redundancy at the controller, AP, power, and software levels. You must know the difference between N+1, N+N, and stateful SSO, and when each is appropriate.
Redundancy models
- N+1: N active controllers share one standby. APs are configured with primary and secondary controllers. On failure, APs detect missed keepalives and rejoin the standby, but client state is lost; failover is typically tens of seconds, not sub-second.
- N+N: each active controller has a dedicated backup, often active/active with primary/secondary assignments. It costs more but avoids a single standby becoming a bottleneck.
- SSO: an active/standby pair synchronizes client and AP state over a redundancy link, giving sub-second failover. On Catalyst 9800, SSO uses a redundancy port and redundancy management interface; both controllers must be the same model and code, and each must be sized to carry the full load after failover.
AP and network redundancy
Configure primary, secondary, and optionally tertiary controllers, and enable AP fallback so APs return to their primary after recovery. Use LAG on the controller for link redundancy, but keep the redundancy port separate. Provide dual power supplies and UPS for controllers and PoE switches. Check licensing: smart licensing must be available on the standby, and AP licenses must cover the full failover load. Plan software upgrades so APs can be upgraded in a rolling fashion and the standby can take over during maintenance.
Worked example
A campus has 800 APs and one C9800-40 rated for 1,000 APs. An N+1 design with one active and one standby works because the standby can carry all 800 APs after failover. If the campus grows to 1,100 APs, the same N+1 pair no longer works: the standby is rated for 1,000 APs, so a single failure would overload it. The fix is to add a second active/standby pair or move to a larger model, and to verify client scale as well as AP scale.
Traps
- Expecting N+1 to preserve client state.
- Mixing controller models or code versions in an SSO pair.
- Sizing each SSO controller only for half the load.
- Forgetting AP fallback or tertiary controller configuration.
- Ignoring PoE and UPS redundancy.
- Overlooking license availability on the standby.