Mobility Support

Mobility is the usually the primary motivation for deploying an 802.11 network. Transmitting data frames while the station is moving will do for data communications what mobile telephony did for voice.

802.11 provides mobility between basic service areas at the link layer. However, it is not aware of anything that happens above the link layer. When designing deploying 802.11, networks engineers must take care so that the seamless transition at the radio layer is also supported at the network protocol layer that the station IP address can be preserved. As far as 802.11 is concerned, there are three types of transitions between access points:

No transition

When stations do not move out of their current access point's service area, no transition is necessary. This state occurs because the station is not moving or it is moving within the basic service area of its current access point.[*] (Arguably, this isn't a transition so much as the absence of a transition, but it is defined in the specification.)

[*] Although my explanation makes it sound as if the "no motion" and "local motion" substates are easily distinguishable, they are not. The underlying physics of RF propagation can make it impossible to tell whether a station is moving because the signal strength can vary with the placement of objects in the room, which, of course, includes the people who may be walking around.

BSS transition

Stations continuously monitor the signal strength and quality from all access points administratively assigned to cover an extended service area. Within an extended service area, 802.11 provides MAC layer mobility. Stations attached to the distribution system can send out frames addressed to the MAC address of a mobile station and let the access points handle the final hop to the mobile station. Distribution system stations do not need to be aware of a mobile station's location as long as it is within the same extended service area.

Figure 2-9 illustrates a BSS transition. The three access points in the picture are all assigned to the same ESS. At the outset, denoted by t=1, the laptop with an 802.11 network card is sitting within AP1's basic service area and is associated with AP1. When the laptop moves out of AP1's basic service area and into AP2's at t=2, a BSS transition occurs. The mobile station uses the reassociation service to associate with AP2, which then starts sending frames to the mobile station.

BSS transitions require the cooperation of access points. In this scenario, AP2 needs to inform AP1 that the mobile station is now associated with AP2. 802.11 does not specify the details of the communications between access points during BSS transitions.

Note that even though two access points are members of the same extended set, they may nonetheless be connected by a router, which is a layer 3 boundary. In such a scenario, there is no way to guarantee seamless connectivity using 802.11 protocols only.

Figure 2-9. BSS transition

ESS transition

An ESS transition refers to the movement from one ESS to a second distinct ESS. 802.11 does not support this type of transition, except to allow the station to associate with an access point in the second ESS once it leaves the first. Higher-layer connections are almost guaranteed to be interrupted. It would be fair to say that 802.11 supports ESS transitions only to the extent that it is relatively easy to attempt associating with an access point in the new extended service area. Maintaining higher-level connections requires support from the protocol suites in question. In the case of TCP/ IP, Mobile IP is required to seamlessly support an ESS transition.

Figure 2-10 illustrates an ESS transition. Four basic service areas are organized into two extended service areas. Seamless transitions from the lefthand ESS to the righthand ESS are not supported. ESS transitions are supported only because the mobile station will quickly associate with an access point in the second ESS. Any active network connections are likely to be dropped when the mobile station leaves the first ESS.

Designing Networks for Mobility

Most networks are designed so that a group of access points provides access to a group of resources. All the access points under control of the networking organization are assigned to the same SSID, and clients are configured to use that SSID when connecting to the wireless network.

As client systems move around, they continuously monitor network connectivity, and shift between access points in the same SSID. 802.11 ensures that clients will be able to move associations between the access points in the same SSID, but network architects must build the network to support mobile clients. Small networks are often built on a single VLAN with a single subnet, in which case there is no need to worry about mobility. Larger networks that span subnet boundaries must apply some additional technology to provide mobility support. Many products can work

Figure 2-10. ESS transition

with a VLAN core, which allows clients to always attach to the same VLAN throughout an organization. New products even allow dynamic VLAN assignment based on authentication data. When users connect to the network, they are attached to the same VLAN everywhere; the switched network simply requires that the wireless LAN device tag frames appropriately. Some products support the Mobile IP standard, or use VPN technology creatively. Trade-offs between all the different mobility strategies are discussed in Chapter 21.

In practice, ESS transitions are quite rare. They usually only occur when users leave one administrative domain for another (say, the corporate network for a hot spot), in which case the two networks in question would have different IP addresses and no trust relationship to transparently attach a client without interrupting network-layer connectivity.

Proprietary mobility systems

Many vendors, especially those who have designed products to build large-scale network environments, have designed their own protocols and procedures for mobility. When I started revising this book, one of the big problems with 802.11 was that it generally required a single IP subnet across an entire roaming area. As a result, wireless LAN deployment often required substantial architecture work as well as major backbone re-engineering efforts. Several vendors rushed to fill the gap by implementing proprietary protocols that enabled sessions to move quickly between access points over arbitrary network topologies. Basic concepts for these solutions will be discussed in the deployment section of this book.