A call that drops while walking and a handheld terminal that freezes between warehouse racks can appear during an access point transition. To understand the cause, check which access point the client is connected to, at what moment it moved, and whether the session was affected at that moment. Adding access points doesn’t always solve this problem. Below we explain what 802.11k, 802.11v and 802.11r actually do during roaming, which to enable when, and how to verify the behavior by measurement. It’s for teams operating enterprise networks that carry voice, video or handheld terminals. If handheld terminals or mobile production equipment are used, plan the field test for industrial Wi-Fi and mobile terminal networks with the real device and route.
Who makes the roaming decision?
In Wi-Fi, the client makes the roaming decision. The AP or controller can’t directly determine which AP the client connects to; it can provide information, make suggestions or force the client to search by cutting the connection. At which signal threshold scanning starts, which AP counts as better and how “aggressively” the transition is made depend on the algorithm of the operating system and wireless card driver. On some drivers, this behavior can be changed with a setting such as “roaming aggressiveness.”
Two practical consequences follow. First, different device models on the same network show different roaming behavior; in a fault investigation, the first question should be “which client, which driver version.” Second, infrastructure settings make the client’s decision easier and faster. They don’t take over the decision. 802.11k, v and r correspond to different stages of this facilitation.
The components of roaming time: scanning, authentication, keys
Splitting the roaming event into three stages makes it clear which standard shortens what:
| Stage | What happens? | Mechanism that shortens it |
|---|---|---|
| Scanning and selection | The client listens on channels or sends probes and finds candidate APs | 802.11k neighbor report |
| Reassociation | Authentication and reassociation frames | 802.11r (moves the key exchange into these frames) |
| Authentication and keys | 802.1X/EAP (on enterprise networks) and the 4-way handshake | 802.11r, PMKSA/OKC caching |
The scanning stage can take longer because of the number of channels at 5 GHz and the requirement for passive listening on DFS channels. On WPA2/WPA3-Enterprise networks, the most expensive step is authentication: if a full EAP exchange is done on every roam, round trips to RADIUS, certificate validation and network latency are added. A data application usually doesn’t let you feel this. Voice does. The common design target in the industry is keeping roaming-induced interruption under 50 milliseconds. For the overall latency budget of voice traffic, ITU-T G.114’s one-way 150 ms recommendation is taken as reference; wireless roaming should consume only a part of this budget.
802.11k: neighbor reports
802.11k (Radio Resource Measurement) defines measurement and report exchange between AP and client. Its most important element for roaming is the Neighbor Report: the client asks the AP it’s connected to for a list of neighboring APs (BSSID, channel and some capability information). That way, instead of scanning all channels, it listens only on the channels in the list and finds a candidate faster.
Its limits:
- The quality of the list depends on the accuracy of the controller’s RF neighbor information. A wrong or incomplete list can steer the client to a bad candidate.
- The client doesn’t have to request the list; whether to use it is up to the driver.
- The neighbor report only shortens scanning time; it has no effect on authentication delay.
The risk of incompatibility between 802.11k and older clients is low. It can be left enabled on most enterprise networks.
802.11v: BSS transition management
802.11v (Wireless Network Management) covers many features; the one that matters for roaming is BSS Transition Management (BTM). The AP sends the client a BTM request containing a candidate list, meaning “you’d be better off moving to these APs.” Typical uses:
- Load balancing: suggesting clients on a busy AP move to a less loaded neighbor.
- Band steering: suggesting a client on 2.4 GHz move to the same AP’s 5 or 6 GHz radio.
- Steering a client with a weak signal to a better AP before the connection drops.
- Moving clients to another AP before planned maintenance (with the “disassociation imminent” flag).
BTM is a suggestion; the client can reject it. If BTM response codes appear in the controller logs, track which device models accept the suggestions. There are few faster ways to find the source of a sticky client problem (a client that stays connected to a distant access point).
802.11r: fast BSS transition and FT modes
802.11r (Fast BSS Transition, FT) combines the authentication and key derivation steps with the reassociation frames. On the first connection, the client does full authentication and the key hierarchy (PMK-R0, PMK-R1) is created. When moving to another AP in the same mobility domain, because the key needed for the new AP is derived in advance, there’s no need to go to RADIUS or do a separate 4-way handshake. That’s exactly where the gain is.
Over-the-air and over-the-DS
FT is done in two ways. In over-the-air mode, the client exchanges FT authentication frames directly with the target AP. In over-the-DS mode, this exchange is relayed to the target AP via the current AP, through the distribution system (the wired network). Over-the-air is more widely supported and more predictable; don’t enable over-the-DS without verifying client and infrastructure support.
The risk of FT incompatibility with older clients
FT is advertised in the SSID’s RSN information as separate AKM (key management) types: FT-802.1X, FT-PSK and, for WPA3-Personal, FT-SAE. Some older drivers that don’t recognize FT can’t connect to the SSID at all when they see this advertisement. We often hear the sentence “we enabled 802.11r and the printers in the warehouse dropped off.” Vendors address this problem in two ways:
- Mixed mode: The SSID advertises both FT and non-FT AKMs; a client that knows FT uses FT, one that doesn’t uses the standard AKM. It’s the recommended starting point on most networks, but problems can still be seen with very old drivers.
- Adaptive FT: some vendors have their own mechanisms that enable FT for specific client families (e.g. Cisco’s Adaptive 802.11r feature). The behavior is vendor-specific; verify it from the documentation.
If FT can’t be used, PMKSA caching or OKC (Opportunistic Key Caching) can be considered to shorten authentication time; these also depend on client support. The interaction of 802.11r with WPA3 is covered separately in the WPA3 migration and transition mode article.
Sticky clients, minimum RSSI and band steering
A sticky client is a client that stays on the AP it first connected to even though its signal has become quite weak. It ruins its own experience; by staying on the air for a long time at a low data rate, it also consumes the capacity of other users in the cell. One bad client slows down the whole cell. The causes are mostly these: the client sees the signal as “sufficient” because AP power is higher than necessary, the client’s roaming threshold is low, and neighboring APs aren’t heard strongly enough at the cell boundary.
Tools on the infrastructure side and their risks:
| Tool | Effect | Risk |
|---|---|---|
| Lowering power and shrinking cells | The client scans earlier | Coverage gaps may form |
| Disabling low data rates | The cell boundary narrows | Older clients can’t connect |
| 802.11v BTM | A gentle suggestion to the client | The client may ignore it |
| Disconnecting with minimum RSSI | Forces the client to search | Uncontrolled disconnect, constant drops at the edge |
| Band steering | Increases 5/6 GHz usage | Delay on first connection, failure on some devices |
Order matters: first the RF design (power and channel width) is fixed, then 802.11k/v are enabled, and forcing tools such as minimum RSSI are used last and in moderation. In networks that do the reverse, users at the edge get disconnected all day. For cell size and power settings, see the channel width and transmit power settings article.
Roaming tests and verification method
The sentence “roaming works” means nothing without measurement. A useful test includes a real walking route and a real client:
- Define the test route: real movement paths such as corridors, floor transitions, the area in front of elevators and between warehouse racks.
- Generate continuous traffic on the client: a low-volume continuous stream for voice (e.g. a VoIP call or a ping at 20 ms intervals), a file copy for data.
- Watch the client on the controller; record the time of roaming events and the source and target APs.
- Measure the interruption time with packet capture or ping losses. On a Linux client,
sudo ping -D -i 0.02 <target-IP>prints each reply with a timestamp; the gap at the moment of roaming is read directly (intervals under 200 ms require administrator privileges on most distributions).
On a Windows client, to produce an HTML report containing wireless connection events, disconnects and reason codes for the last three days, the following command is used at an administrator command prompt; the report is written to C:\ProgramData\Microsoft\Windows\WlanReport\wlan-report-latest.html:
netsh wlan show wlanreport
netsh wlan show interfaces
Wireless packet capture requires an adapter running in monitor mode on the client’s channel. Putting the adapter into monitor mode on Linux:
sudo ip link set wlan1 down
sudo iw dev wlan1 set type monitor
sudo ip link set wlan1 up
sudo iw dev wlan1 set channel 36
Useful display filters for examining the moment of roaming in Wireshark:
| Filter | What it shows |
|---|---|
wlan.fc.type_subtype == 0x0b |
Authentication frames |
wlan.fc.type_subtype == 0x02 |
Reassociation request |
wlan.fc.type_subtype == 0x03 |
Reassociation response |
eapol |
EAP and the 4-way handshake |
wlan.fixed.auth.alg == 2 |
FT authentication |
If FT is working, the algorithm value in the authentication frame shows as FT, and no EAP or 4-way handshake frames are seen after reassociation. If you see a full EAP exchange, the client isn’t using FT. The evidence is in the packets. Because capture is done on a single channel, if the target AP is on a different channel you need to listen on that channel too with a second adapter. For the general method of packet capture, you can use the network packet analysis with Wireshark article; the primary source for the standard’s definitions is the IEEE 802.11 working group.
On networks with voice and handheld terminals, write the roaming test into the installation acceptance criteria; it greatly reduces post-installation surprises. Projects where roaming measurement and RF settings need to be done together fall within the wireless network installation and optimization service.
Frequently asked questions
Does enabling 802.11r cause problems with older clients?
It can. Some older drivers that don’t recognize the FT AKM can’t connect, or behave unstably, when FT is advertised in the SSID’s RSN information. Test FT first on a pilot SSID with the problematic device models in the inventory; if needed, use mixed mode (where FT and non-FT AKMs are advertised together) or a separate SSID.
How long should roaming take on enterprise Wi-Fi?
Data applications usually tolerate an interruption of a few hundred milliseconds. For voice and video calls, the common design target in the industry is keeping roaming-induced interruption under 50 milliseconds. Full 802.1X authentication mostly exceeds this target; that’s why FT or key caching is needed.
Can an AP force a client to move to another AP with 802.11v?
No, 802.11v BSS transition management is a suggestion. The client can accept, reject or ignore the request. With the “disassociation imminent” flag, the AP can announce that it will cut the connection after a certain time; even then, the client chooses which AP to move to.
Does a minimum RSSI setting solve roaming problems?
To a limited extent. The AP forces a client whose signal falls below the threshold to look for a new AP by cutting its connection; however, cutting the connection doesn’t guarantee the client will move to a particular AP; the client may try to reconnect to the same AP. This is an uncontrolled disconnect, and if the threshold is chosen wrongly, users at the edge of coverage keep getting disconnected. Cell size and power settings should be fixed first, and minimum RSSI used as a last resort with a careful threshold.