Going Wi-Fi: A Practical Guide to Planning and Building an 802.11 Network

The fundamental components of a WLAN are clear-cut. And it's relatively easy to describe the 802.11 suite in basic networking terms-a series of specifications that provide an open asynchronous networking environment that requires a distributed control function. But the technology isn't as simple as it sounds.

Wired LANs use a physical medium to interconnect their terminals. Access nodes are provided at various points on the physical medium to allow for the connection of terminals to the medium. LANs may be interconnected by means of bridges or switches.

The most popular set of LAN specifications is 802.3 (commonly referred to as "Ethernet") with all of its evolutionary variations. These specifications provide for message structuring rules, station naming, allocation of resources, and other housekeeping functions.

Most wireless LANs extend their access to a wired LAN's medium via an Access Point (AP) that attaches as a bridge to the wired LAN. (I say most because you can set up a WLAN that stands alone, i.e. no wired access whatsoever.) The AP uses radio spectrum to extend the LAN's medium to radio equipped network devices within the AP's range.

In a mixed wired/wireless environment, the most common WLAN configuration uses a single AP to provide service to all terminals within its coverage area. In such situations, the AP is analogous to a wired network's hub in that it supports the shared usage of the medium by its active computing devices. A wireless computing device becomes a full-fledged member of the wired LAN, with all assigned LAN privileges, after it is associated with an AP. Wireless stations (e.g. computing devices) may talk to each other, whether they are on the same or different APs, and they can also communicate with devices on the wired LAN.

Mobile computing devices may roam among multiple access points without losing their connection. However, if the access points are in separate subnets within the wired LAN and there is no specific roaming technology in residence, it may not be possible for a station to maintain session persistence when roaming outside its subnet due to routing constraints within the LAN.

The 802.11 series of standards that are spreading throughout today's networking landscape are sometimes being described as wireless Ethernet. This is because they use similar modulation techniques and are simple and flexible. And perhaps also because Bob Metcalfe (he is considered the "inventor" of Ethernet) has stated that he chose the word "Ether" because he didn't want Ethernet to be associated with a particular media. So there is some logic behind the term wireless Ethernet.

But although the 802.11 standards share some common aspects of an Ethernet network, in reality, the 802.11 specifications provide for a CSMA/CA (Carrier Sense Multiple Access/Collision Avoidance) network while 802.3 provides for a CSMA/CD (Carrier Sense Multiple Access/Collision Detection) network. To understand, let's consider the more established 802.11b standard since it is still the most widely implemented WLAN technology.

The 802.11b specification is designed to use a variant of Ethernet's CSMA/CD-CSMA/CA. CSMA/CA was chosen because CSMA/CD would require that the wireless radios be able to send and receive at the same time. Not only would that serve to substantially increase 802.11b product price and complexity, but also, in a wireless networking environment devices are not always in a position to hear all of the other wireless devices on the network.

CSMA/CA utilizes the RTS/CTS (request to send/clear to send) protocol to notify other workstations that a transmission is about to take place. This four-way handshaking minimizes the number of collisions and makes sure that hidden nodes are aware of transmissions across the entire wireless segment; however, this method introduces significant overhead on the network. 802.11's MAC is also significantly more complex than a typical 802.3 MAC, because the wireless specification calls for four MAC addresses instead of the two found in an Ethernet header.

Finally, to maintain backward compatibility with the original 802.11 specification devices, 802.11b wireless device transmits the preamble and a portion of the packet header at 1 Mbps. This accounts for significant additional overhead, as that preamble is significantly longer than an Ethernet preamble. The overall result is a network that can be at most 70 percent efficient (allowing a maximum data throughput rate of about 7.7 Mbps).

Losses in effective transmission rates can be reduced by improving the strength of the primary signal, which, in turn, reduces the time it takes to discern ghost signals from the true signal and the amount of time it takes to sample diverse antennae. Furthermore, many of the 802.11b WLAN products on the market today push throughput at the cost of interoperability. In the case of point-to-point devices, eliminating the 1 Mbps 802.11 legacy transmit speed can significantly improve performance. Likewise, if a network is known to be point-to-point, the Random Backoff algorithm, interframe gap, and preamble can all be minimized to maximize throughput.

This adds up to a wireless standard that really isn't Ethernet at all. When Xerox first licensed Ethernet, it charged a pittance, but in exchange it stipulated that the technology couldn't be changed; it had to interoperate with all other Ethernet implementations. The 802.11 series doesn't meet that directive.