Last month, we described the qualities of a structured wireless network. This month, we continue that discussion with some specific examples of technology that can be used to create a structured wireless network. As we do so, it’s important to realize that this term doesn’t have a single, standard definition. For example, the term is used in a general sense to describe networks like switched Ethernet networks, while at the same time, Cisco uses the term to describe a wireless architecture using specific Cisco components and protocols. The connotation of the term “structured wireless networking” depends very much on the context in which that term is used. But regardless of the context, all uses of the term seem to have the major characteristics that we describe in this document: easy deployment with minimal pre-deployment planning, easy expansion of the network without major reconfiguration of un-affected parts, and centralized management.

Cisco’s SWAN

Cisco’s “Structured Wireless-Aware Network” (SWAN) is one example of a structured wireless network. A SWAN consists of various Cisco components (access points, routers, switches, management consoles, and clients) working together to implement a structured wireless network. Cisco’s approach is to use software upgrades in traditional wired devices, such as switches and routers, to more tightly integrate the wireless network into the existing wired network. SWAN provides centralized authentication, easy addition of access points to the network, fast roaming between APs (this is useful for Voice Over Wi-Fi, where the delay of normal roaming might create audible distortion of the call), automatic rogue AP detection, and assisted and automatic site surveys. One caveat of Cisco’s approach is that it only works if you own Cisco equipment. A customer that uses non-Cisco equipment or who has a mixed network can’t take full advantage of SWAN.

The Evolution of Structured Wired and Wireless Networks

Just as wired Ethernet networks have evolved from less structured to more structured, so are wireless networks. “Physical architecture” refers to the way in which the wires connect the stations to each other. “Logical architecture” refers to the way in which packets move across the wires. In general, a network whose logical architecture is separate from its physical architecture is more conducive to structured design. Just as wired Ethernet evolved from a network whose physical architecture was closely tied to the logical architecture, so are wireless networks.

Ethernet started with both a physical and a logical bus architecture. Physically, a single wire ran between all of the stations (this is the definition of a “bus”). Logically, when one station transmitted a packet, the packet traveled down the wire and all stations that were attached to the wire heard the packet. The flow of packets (logical architecture) was dictated by the electrical characteristics of the wire (the physical architecture).The only way to change the flow of packets was to change the wire itself (move some stations to a different wire, for example). When Ethernet transitioned to 10BaseT, it ceased to be a physical bus; all stations were star-wired back to a hub. But logically, 10BaseT was still a bus architecture. When one station transmitted a packet, the hub repeated the packet to all other stations so that they heard the packet, just as they would have in a physical bus.

Because 10BaseT separated the pattern of traffic flow (logical architecture) from the layout of the wires (physical architecture), it made the network easier to deploy and expand—both qualities of a structured network. 10BaseT was, therefore, more conducive to structured design than 10Base5 or 10Base2. How does this observation apply to wireless networks, then? In a wireless network, so far, the physical architecture is always a shared bus. One station transmits and all other stations within range “hear” the transmission. The only method of manipulating that architecture is to place access points on separate channels. Since only three non-overlapping channels are available in 802.11b and 802.11g, options for manipulating the physical architecture are limited.

Limited options for manipulating the physical architecture are one reason why most 802.11 networks are less structured today. But, just as Ethernet evolved more sophisticated physical architectures, so is 802.11. For example, some vendors make “sectorized” antennas that are capable of beaming signal to more than one station on the same channel at the same time without interference (this technology is currently still in the early stages of development). This is roughly analogous to the shift in Ethernet from coaxial bus wiring to switched twisted-pair wiring.

Although 802.11’s physical architecture is essentially always a shared bus, many options exist to manipulate the logical architecture, or packet flow. In a wireless distribution system, access points forward packets from one to the other, similar to the way that routers or switches forward packets. But a wireless distribution system is static and manually-configured, which isn’t very structured.

A routed mesh is a far more structured way of manipulating the logical architecture of a wireless network. In a routed mesh, the wireless routers automatically detect each other and configure forwarding paths between themselves. If one router fails, the remaining devices will route around it, if possible. Notice that this scenario is more structured because it has automatic configuration, which leads to easy deployment and scalability. This scenario also separates the physical and logical architectures and gives the administrator more control over the logical architecture. That these two qualities are found together is not a coincidence. This is an example of a network technology that separates the logical architecture (packet flow) from the physical architecture and increases its potential for structured design as a result. A wireless switch is another common 802.11 technology that achieves a structured nature by separating the physical architecture from the logical flow of packets.

Applying Structured Wireless Design

Developing a structured design requires appropriately integrating various architectural approaches to the overall network infrastructure. Two types of infrastructure include:

• Point to Point (PtP) – An infrastructure in which a single link carries information between two areas of connectivity, such as when a WAN link connects two separate campuses of a single organization.
• Point to Multipoint (PtMP) – An infrastructure in which many outlying regions of connectivity link back to one central region of connectivity, such as when campus buildings link back to a central building that has an Internet uplink.

Methods of constructing the architecture include:

• Wireless Mesh – A network topology that can contain multiple paths between two nodes. In the extreme, each node has a dedicated connection to each other node. In reality, this is not usually possible, and real-world meshes usually manifest some subset of this topology—i.e. each node is connected to all other nodes that are within range of it. Wireless meshes usually have the characteristic that they are able to dynamically rearrange themselves in response to downed links.
• Wireless Tree – A network topology that contains only one unique path between any two nodes. This makes the topology simpler from the perspective of the access points or routers, but it also means that a single failed link can isolate parts of the network, since a tree cannot have redundant, backup links. Wireless tree topologies are usually implemented using access points configured in a Wireless Distribution System (WDS). A WDS differs from a wireless mesh in that the links are usually manually configured in a WDS, while they are automatically configured in a mesh, and that a WDS cannot route around failed links because it cannot have redundant or backup links at all.

From left to right, examples of: Point to Point (as would be made with WDS), Point to Multipoint (as would be made with WDS), and mesh topologies.

These two sets of concepts interrelate. The first two (PtP and PtMP) describe the topology of the network’s links. The second two (mesh and WDS) describe the manner in which the topology is configured. An appropriately structured wireless network that meets the objectives of manageability and scalability identifies and specifies architectural implementations that are consistent with the application, bandwidth, and geographic requirements of the system.

To illustrate that point, we have constructed an example of a structured wireless network that involves all of the architectural concepts that we have introduced so far. In this example, a campus of multi-story apartment buildings requires wireless connectivity to the Internet. Because the buildings are located in an area with a poor local phone infrastructure, Internet access must be brought in wirelessly. (We have seen this situation in some cities in Africa, where the wired telephone infrastructure is underdeveloped and cellular phones predominate).

The first component of the infrastructure, then, involves getting Internet access from a large city nearby to the smaller city with the apartment buildings. This distance might be as much as ten or twenty miles, depending on the equipment used and the terrain between the cities. Since only two cities are involved here, this will be a point-to-point link. A point-to-point link wouldn’t benefit from being configured as a wireless mesh, since a wireless mesh only manifests its benefits when there are many nodes and many paths between the nodes. Therefore, it would be best to configure this link as a WDS. When a WDS is configured between just two nodes, in a point-to-point configuration, it is often referred to as a wireless bridge link.

Next, we must determine where the terminating point of the point-to-point link that brings in the Internet access will be. Perhaps we decide to use a particularly tall building in the city, since the taller building gives more clearance over earth bulge and any terrain obstructions. We place a directional antenna on that building pointing back towards the large city form which we are drawing Internet access. Next, we must distribute that connectivity to the individual apartment buildings. In this case, we are extending connectivity from a single point to multiple points, meaning that this will be a point-to-multipoint link.

Whether this should be done using mesh or WDS depends on the physical layout of the apartment buildings in question. If they are close enough together that a mesh router on one could see a mesh router on the others, then they might benefit from mesh routing’s ability to route around problems. If they are far enough apart that they couldn’t see each other, then WDS would be the best choice.

Once signal has reached an individual apartment building, we have to distribute it within the building. Typically, wireless signal propagates relatively freely through the floors of an apartment building, but the concrete that is usually laid between floors prevents signal from going between floors. A structured way of addressing this situation would be to run signal through an elevator shaft or stairwell and then to distribute the signal to each floor individually. To do this, a wireless device is placed on each floor in the elevator shaft or stairwell. Although this could be done with a WDS, a more structured design would be to use mesh routers instead. Remember that a WDS is statically configured and cannot route around failures. Therefore, if a single floor were to go down for some reason, it would break the chain of connectivity back to the roof and all floors below that would lose Internet access. A mesh router could easily route around that failure.

Within a single floor, regular access points could be used, which would provide client access to the network. These access points could be connected back to the mesh router in the stairwell or elevator shaft with a WDS if it was impractical to run Ethernet cabling. A diagram of this type of network is shown below.

Lightning Protection

I’m installing an outdoor antenna. I know that I need a lightning arrestor, but I’m not sure how to install it. Can you help?

Lightning protection is essential for all outdoor antenna installations. Without it, a nearby lightning strike can damage expensive equipment and cause injuries. A lightning arrestor is a device that installs between the antenna and the access point. If a lightning strike occurs, the arrestor shunts the transient current induced by the strike off to ground.

The first rule when installing a lightning arrestor is that it must have a good earth ground. An ungrounded arrestor has nowhere to send the transient current from the strike. A lightning arrestor will have a grounding screw that must be connected to the ground, be that a grounding rod driven into the earth or the metal frame of the building. Importantly, you should not ground the arrestor to your building’s electrical system (e.g. the third prong on an A/C plug). Although the electrical system does provide a good earth ground, grounding the arrestor to the electrical system will shunt dangerous current from the antenna into the electrical system. The AP might survive, but anything plugged into the electrical system might not.

The arrestor is usually screwed directly into the antenna connector on the access point, bridge, or other radio. This also allows it to perform double-duty as a converter between the radio’s proprietary RF connector (RP-TNC, SMA, etc…) and the standard N-connector that is probably found on the cables. If you use the arrestor in this way, make sure to order one with the correct connectors for the equipment you’re using.

If installing the arrestor near the AP is not convenient, it can be installed at the antenna-side of the cable, although this often results in more difficulty in finding a ground. Keep in mind that only equipment on the radio-side of the arrestor will be protected, so if you have an expensive amplifier, you might want to put the arrestor between the amplifier and the antenna, as opposed to between the radio and the amplifier. Another implication of this principle is that the antenna is always exposed to lightning damage and may need to be replaced in the event of a strike. Installing lightning rods near to the antenna might provide some protection if the antenna itself is particularly valuable.

Lightning protection is one area where you don’t want to take any chances. Grounding and A/C current flow can be complex. In some cases, it might make sense to hire a professional electrician or an experienced outdoor antenna installer.

TYPICAL LIGHTNING ARRESTOR