Wireless RRM and Roaming: DCA, TPC, Mobility Groups, and Rogue Detection

Radio Resource Management (RRM) is the set of algorithms a Cisco wireless LAN controller runs to keep a group of access points tuned without an engineer choosing settings by hand. Two of them carry most of the exam weight: Dynamic Channel Assignment (DCA) decides which channel each radio uses, and Transmit Power Control (TPC) decides how loudly each radio talks. Roaming is separate. Clients move between access points on their own, and mobility groups let a client keep its session when the next access point belongs to a different controller. Wireless has no domain of its own on the blueprint, which is part of why it is one of the areas candidates routinely under-study while over-preparing routing and switching.

RF groups

RRM does not operate per access point. Controllers that share an RF group name and hear each other’s access points form an RF group, and one controller is elected the RF group leader. The leader collects the measurements every access point reports, runs the algorithms across the set, and pushes the resulting channel and power assignments back out, because a channel change at one access point only makes sense in the context of its neighbors. Access points build those neighbor lists by hearing each other’s discovery messages over the air, and the signal strength of those messages is the raw input both algorithms consume.

Dynamic Channel Assignment

DCA assigns channels so that nearby cells do not sit on top of one another. Its inputs are what each radio hears: neighboring access points on each channel, co-channel interference, noise from non-Wi-Fi sources such as microwave ovens, and channel load.

It runs on an interval, ten minutes by default, and evaluates whether a change would lower the overall cost across the group. Because moving a radio disrupts the clients on it, DCA acts only when the improvement clears a sensitivity threshold. A startup mode runs aggressively for a period after a large change, which is the right tool after installing a batch of access points.

The condition DCA exists to fix is the one produced by adding access points to a dense area: several radios settle onto overlapping 2.4 GHz channels and throughput falls across the floor. Only channels 1, 6, and 11 are non-overlapping there in most regulatory domains, and a dozen radios cannot all get a clean one.

Transmit Power Control

TPC works in the other dimension, and its trigger is an access point hearing its neighbors too strongly. The algorithm examines the signal level at which each access point hears the third-loudest neighbor above a configured threshold. If that exceeds what the design intends, the cell is larger than necessary and the radio’s transmit power is reduced. A smaller cell overlaps less into adjacent cells, which cuts co-channel interference.

Power is only lowered by this path. Raising power is the job of coverage hole detection and correction, which watches for clients reporting poor signal quality and increases an access point’s power to reach them. Applying that to a room whose cells already overlap would worsen the interference, which is why the two are separate features triggered by different evidence.

	Dynamic Channel Assignment	Transmit Power Control
Changes	The radio’s channel	The radio’s transmit power
Input	Interference, noise, load	Third-loudest neighbor’s signal
Trigger	Overlapping or congested channels	Neighbors above the threshold
Direction	Lateral, to a better channel	Downward, to a smaller cell

Two related features appear as wrong answers. Band select nudges dual-band clients toward 5 GHz by delaying probe responses on 2.4 GHz, and changes no channel or power. Regulatory domains are set by national regulators and encoded in the platform, so no algorithm creates them.

Machine learning in radio tuning

Recent Cisco wireless platforms extend RRM by feeding stored radio telemetry into machine learning models rather than deciding from the current snapshot alone. The system learns which channel and power settings produced good outcomes under particular interference conditions and applies that history to future assignments, predicting a recurring pattern before it degrades service. This is the AI and machine learning material added to the 200-301 v1.1 blueprint, and the discriminator is learning from stored data. A static channel plan never learns, and an API call returning an access point inventory analyzes nothing.

Roaming

A client, not the infrastructure, decides when to roam. It monitors the signal from its current access point, scans for alternatives, and reassociates when a better candidate appears. The infrastructure controls only whether that move costs the client its session.

Intra-controller roaming happens when both access points are joined to the same controller. That controller already holds the client’s context, including its security association and IP address, and moves the association across.

Inter-controller roaming happens when the two access points belong to different controllers. The new controller has no context for the client and must obtain it from the old one. When both place the WLAN on an interface in the same VLAN and subnet, the context is handed over and the client keeps its address. When they serve different subnets, the original controller becomes the anchor and the new one the foreign controller, and traffic is tunneled back to the anchor so the address stays valid.

Mobility groups

Controllers exchange client context only with peers they have been told about and that claim the same group. Two things must be configured on both:

• The same mobility group name, the membership test. A mismatch, including a difference in case, prevents the handoff.
• A mobility peer list entry for the other controller, carrying its management IP address and MAC address, which is how each device reaches the other.

Each controller keeps its own distinct management address, so any answer proposing that two controllers share one management IP is wrong on its face. Mobility messaging is routed between management interfaces and needs no direct cable. The service port is an out-of-band administrative connection and plays no part in mobility, a distinction laid out in WLC ports and interfaces.

Rogue detection

Access points serving clients spend brief intervals off-channel listening, and a monitor mode access point does nothing but listen, so the controller accumulates a list of every basic service set and client MAC address heard in the air. Being nearby is not by itself actionable: a neighboring tenant’s network is heard and is not a threat.

The consequential question is whether an unauthorized device is attached to the wired network, and a rogue detector access point answers it. Its radios are switched off and it works from the wire, receiving from the controller the MAC addresses heard over the air and watching wired traffic for those addresses. A match means the device is plugged into the network, not merely in range. Access point roles are compared in the article on Cisco AP modes.

Rogue containment is a separate action in which access points transmit deauthentication frames to break associations to the rogue. It requires working radios, so a rogue detector cannot perform it, and it carries legal exposure when the target belongs to a neighbor.

How the 200-301 exam tests this

• Which algorithm fixes this symptom. Overlapping channels after new access points are added is DCA. Cells overlapping because radios hear each other loudly is TPC, which lowers power. Clients with weak signal is coverage hole correction, which raises it. Steering clients to 5 GHz is band select.
• Roaming prerequisites. A multi-select asks what must be configured on both controllers: the group name and the peer list entries. Shared management addresses, a cable between controllers, and service-port settings are the distractors.
• What a rogue detector does. It correlates over-the-air MAC addresses against wired traffic with its radios off. Associating to the rogue, deauthenticating it, or triangulating it describe other roles.
• Recognizing learned behavior. A platform improving assignments from historical telemetry is machine learning applied to RRM, not a regulatory function, a static plan, or an inventory query.

Symptom-to-algorithm mapping is quick to test — a block of practice questions tells you whether DCA and TPC have fully separated in your head.

Quick reference

• DCA changes channels; TPC changes transmit power.
• TPC lowers power based on the third-loudest neighbor above the threshold.
• Coverage hole detection and correction is the feature that raises power.
• Band select influences client band choice and changes no radio setting.
• Inter-controller roaming needs a matching group name plus mutual peer entries with management IP and MAC.
• Different subnets produce an anchor and foreign pair with traffic tunneled to the anchor.
• A rogue detector AP has its radios off and matches air-heard MACs against wired traffic.