Roaming is the method where client stations move between RF coverage cells in a seamless manner. Client stations switch communications through different access points.
Seamless communications for stations moving between the coverage zones within an Extended Service Set (ESS) is vital for uninterrupted mobility. One of the most common issues you’ll need troubleshoot is problems with roaming.
Roaming problems are usually caused by poor network design. Due to the proprietary nature of roaming, problems can also occur when radio cards from multiple vendors are deployed.
Changes in the WLAN environment can also cause roaming hiccups. Client stations and not the access point make the decision on whether or not to roam between access points.
Some vendors may involve the access point or wireless switch in the roaming decision, but ultimately, the client station initiates the roaming process with a re-association request frame.
The method in which client stations decide how to roam is entirely proprietary. All vendor client stations use roaming algorithms that can be based on multiple variables.
The variable of most importance will always be received signal strength. As the received signal from the original AP grows weaker and a station hears a stronger signal from another known access point, the station will initiate the roaming process.
However, other variables such as SNR, error rates, and retransmissions may also have a part in the roaming decision. Because roaming is proprietary, a specific vendor client station may roam sooner than a second vendor client station as they move through various coverage cells.
Some vendors like to encourage roaming while others use algorithms that roam at lower received signal thresholds. In an environment where a WLAN administrator must support multiple vendor radios, different roaming behaviors will most assuredly be seen.
For the time being, a WLAN administrator will always face unique challenges because of the proprietary nature of roaming. In the future, the 802.11k draft and much anticipated 802.11r roaming draft will hopefully standardize many aspects of roaming.
The best way to assure that seamless roaming will commence is proper design and a thorough site survey. When designing an 802.11 WLAN, most vendors recommend 15 to 20 percent overlap in coverage cells at the lowest desired signal level.
The only way to determine if proper cell overlap is in place is by conducting a coverage analysis site survey. Roaming problems will occur if there is not enough overlap in cell coverage.
Too little overlap will effectively create a roaming dead zone, and connectivity may even temporarily be lost. On the flip side, too much cell overlap will also cause roaming problems.
For example, if two cells have 60 percent overlap, a station may stay associated with its original AP and not connect to a second access point even though the station is directly underneath the second access point.
This can also create a situation in which the client device is constantly switching back and forth between the two or more APs. This often presents itself when a client device is directly under an AP and there are constant dropped frames.
Another design issue of great importance is latency. The 802.11i amendment defines an 802.1X/EAP security solution in the enterprise. The average time involved during the authentication process can be 700 milliseconds or longer.
Every time a client station roams to a new access point, reauthentication is required when an 802.1X/EAP security solution has been deployed. The time delay that is a result of the authentication process can cause serious interruptions with time-sensitive applications.
VoWiFi requires a handoff of 50 milliseconds or less when roaming. A fast secure roaming (FSR) solution is needed if 802.1X/EAP security and time-sensitive applications are used together in a wireless network.
Currently, FSR solutions are proprietary, although the 802.11i amendment defines optional FSR and the 802.11r draft will hopefully standardize fast secure roaming. Changes in the WLAN environment can also cause roaming headaches.
RF interference will always affect the performance of a wireless network and can make roaming problematic as well. Very often new construction in a building will affect the coverage of a WLAN.
If the physical environment where the WLAN is deployed changes, the coverage design may have to change as well. It is always a good idea to periodically conduct a coverage survey to monitor changes in coverage patterns.
Layer 3 Roaming
One major consideration when designing a WLAN is what happens when client stations roam across layer 3 boundaries. As pictured in Figure below, the client station is roaming between two access points.
The roam is seamless at layer 2, but a router sits between the two access points and each access point resides in a separate subnet. In other words, the client station will lose layer 3 connectivity and must acquire a new IP address.
Any connection oriented applications that are running when the client reestablishes layer 3 connectivity will have to be restarted. For example, a VoIP phone conversation would disconnect in this scenario and the call would have to be reestablished.
The preferred method when designing a WLAN is to only have overlapping Wi-Fi cells that exist in the same layer 3 domains through the use of VLANs.
However, because 802.11 wireless networks are usually integrated into preexisting wired topologies, crossing layer 3 boundaries is often a necessity, especially in large deployments.
The only way to maintain upper-layer communications when crossing layer 3 subnets is to provide either a Mobile IP solution or a proprietary layer 3 roaming solution.
Mobile IP is an Internet Engineering Task Force (IETF) standard protocol that allows mobile device users to move from one layer 3 network to another while maintaining their original IP address. Mobile IP is defined in IETF request for comment (RFC) 3344.
Mobile IP and proprietary solutions both use some type of tunneling method and IP header encapsulation to allow packets to traverse between separate layer 3 domains with the goal of maintaining upper-layer communications.
Most wireless switches and controllers now support some type of layer 3 roaming solution. While maintaining upper-layer connectivity is possible with these layer 3 roaming solutions, increased latency is often an issue.
Additionally, it may not be a requirement for your network. Even if there are layer 3 boundaries, your users may not need to seamlessly roam between subnets. Before you go to all the hassle of building a roaming solution, be sure to properly define your requirements.