Centralised approaches require a tunnelling protocol between the access points and the central or distributed controllers. Proposed standards include Light Weight Access Point Protocol. Some centralised approaches use Generic Routing Encapsulation tunnelling, with special extensions to deal with roaming issues. All of these solutions should be considered proprietary, and will only become an issue if mixed controller vendors are involved. Security models can vary among vendors, with some using VPN technology that is enhanced to roam. Others provide their own methods derived from standards such as WPA. This overlay of a security mechanism can be effective -- in some cases, enabling access points to be intermixed. This approach is particularly useful when several access-point brands have been deployed and the IS organisation is attempting to consolidate systems into a manageable environment. Universities are good examples of this type of WLAN deployment; many universities install high-capacity centralised controllers that overlay a security, management and functionality framework on top of the established infrastructure. (The popularity of this deployment method may increase as more organisations separate basic IP connectivity from the capabilities that will be deployed as overlays.)
Large offices -- In large offices, access points will be connected to local switches for network connectivity, as well as for power over Ethernet (POE). The architectures we have discussed can be employed, although in the long term it will probably make sense for new switches to be purchased with WLAN support. POE will also be required for voice over IP (VoIP) deployment. High bandwidth and capacity support issues may also need to be addressed in large offices. Although the initial deployment may be at 802.11b, 802.11a may be needed as a complement to support greater throughput and capacity. As many as 24 channels are available in 802.11a, and access points can be stacked to improve capacity. If some areas require minimum 54Mbps link rates, 802.11h capabilities can be used to reduce cell sizes, permitting the deployment of cells at close inter-access-point distances to maintain the 54Mbps rate. With some access points, a specialty 108Mbps rate is available, but this is not standard. Should all else fail, wire must be deployed, but this should prove to be the exception. Large offices require effective planning through site surveys of ongoing best management practices, including periodic reviews.
Physical-layer choices:
Enterprises must decide whether to deploy Institute of Electrical and Electronics Engineers (IEEE) 802.11a, 802.11b, 802.11g (which includes support for 802.11b) or a combination of these, and which frequencies to use.
802.11a -- Choosing 802.11a depends on the maturity of the technology. 802.11a operates at several 5GHz frequency bands and has the potential to supply as many as 24 channels at a 54Mbps link rate. However 802.11a silicon has matured and is awaiting the approval of some important related standards -- namely, 802.11h, which supports automatic power control and dynamic frequency selection for European approval. Also various countries have been settling regulatory issues so that silicon manufacturers can develop more-generic worldwide technology. These technologies should ship as an integrated whole by year-end 2004. When this occurs, we will shift our recommendation from 802.11b as the primary delivery technology to 802.11a. At that time, we also expect that WPA2 (an interpretation of 802.11i with Advanced Encryption Standard [AES] encryption tested for interoperability by the Wi-Fi Alliance) will come to bear, making the combination with 802.11a less costly and more attractive than separate deployments (WPA/WPA2/802.11i may not be appropriate for all applications).
