Wi-Fi® involves over-the-air communications in two unlicensed frequency bands, the 2.4 GHz and 5 GHz bands, that are available worldwide. 802.11b and 802.11g operate only in the 2.4 GHz band. 802.11a operates in the 5 GHz band, while 802.11n can operate in both bands.
The vast majority of Wi-Fi client devices operate only in the 2.4 GHz band, which has only three non-overlapping channels. The 2.4 GHz band also is the home of Bluetooth® devices, cordless phones, microwave ovens, and other devices. The 5 GHz band is less cluttered and offers considerably more capacity, with up to 23 non-overlapping channels available (depending on regulatory domain). With more channels and fewer sources of RF interference, the 5 GHz band is attractive for a hospital Wi-Fi deployment, especially when the deployment covers many floors and APs are relatively dense.
There are two types of 5 GHz channels: those for which dynamic frequency selection is required (DFS channels), and those for which DFS is not required (non-DFS channels). DFS, a mechanism defined in IEEE 802.11h, ensures that Wi-Fi transmissions will not interfere with radar. Using DFS channels with highly mobile devices may be problematic, depending on how those devices handle roaming.
A Wi-Fi client connects to an enterprise network through an infrastructure endpoint called an access point (AP). When the client moves to a position where its connection to that AP becomes suboptimal, the client will try to switch to an AP that provides better connectivity. The process of switching from one AP to another is called roaming. The client must disconnect from the AP to which it is associated before initiating a connection to another AP. The longer the time between the disconnection from the network and the reconnection to the network, the greater the loss of packets and disruption to the applications running on the device.
The decision of when to roam, the selection of the target AP, and the management of the roaming process are handled by the client device’s Wi-Fi radio software. The roaming process has four steps:
- Evaluate: Determine if the connection with the current AP is less than optimal.
- Scan: If the connection is less than optimal, then scan the vicinity for other APs.
- Select: Determine if any AP within range is likely to provide a better connection than the current AP.
- Roam: Disconnect from the current AP and create a connection to the best candidate in the vicinity and authenticate to the network.
Scanning for available APs is disruptive to network connectivity, because while a client radio scans it is “off channel”, or not communicating with its current AP. Some client radios scan only when the current connection is suboptimal. Others do periodic scans, when the radio is not sending or receiving data to the current AP, and cache the results of each scan for a period of time that may be configurable. For a highly mobile device, the APs that are available when a roam is required may differ from the APs that are listed in the cached information. When a client’s applications require a persistent network connection, the optimal approach is to perform the scan just before the roam.
The larger number of channels available in the 5 GHz band provides for increased network capacity, but it also can lead to longer roam times for mobile devices, because effective roaming requires scanning of every available channel. When available channels include those that require DFS, scan times are even longer, because DFS channels require passive scanning, which is slower than active scanning. Network administrators should reserve DFS channels for non-mobile devices or look for Wi-Fi clients that support optimized DFS channel scanning capabilities.