Choosing Your Logical Architecture

When choosing a logical architecture, you must weigh several trade-offs. Some of these are security trade-offs are discussed in the next chapter. Many are, however, a matter of balancing performance, simplicity, or functionality.

1. Mobility is a baseline function that should be provided by any architecture you choose. 802.11 provides for mobility within an extended service set, and that ESS must be visible to the client as a single IP subnet. All of the architectures presented here attach clients to a single subnet, although the mechanics of how they do so differ radically.
   1. For small-scale deployments using a handful of APs, any of the architectures work. The first two are easy to set up on a very small scale, and may have the advantage for cost-conscious deployments that will never grow beyond the initial handful of APs.
   2. The IEEE's inter-access point protocol provides link-layer mobility only. Crossing router boundaries into new broadcast domains requires network-layer coordination between wireless LAN access devices. At the time this book was written, subnet mobility generally required picking a single-vendor solution. Mobile IP is an open standard, but it is not widely implemented.

2. Clients must perceive that they are attached to a single IP subnet, no matter what the physical location of their attachment. This does not, however, require that all clients be attached to the same subnet. Multiple subnets may overlap "over the air." Multiple subnets over the air offer the ability to more finely control user access privileges and differentiate between user groups, but require the use of 802.1X.
3. What limitations does the existing network impose? Baggage from past decisions may limit the choices that you can make.
   1. A sprawling network with a large diameter may not be able to extend VLANs across the entire network due to spanning tree limitations. This may rule out the use of a single wireless VLAN, or a dynamic VLAN model where the access points must be connected to the core.
   2. The dynamic VLAN topology may depend on widely distributing 802.1Q tags throughout your network. If VLAN information is not already available, network administrators must find a way to distribute it to all the locations that support a wireless network. Products that require direct connection into the core are incompatible with a routed core network.
   3. Networks may have choke points in a variety of places. Pre-existing choke points may limit the number of wireless devices that can be attached in many locations. If your desired architecture intentionally introduces a choke point, it must be fast enough to not limit throughput.

4. The choice of network topology may be driven in part by the security protocols used on the network. Dynamic VLAN assignment is possible only with 802.1X, so the last two topologies work best for administrators who want to use link layer security mechanisms. The first two topologies are much more suited to use with network-layer security based on IPsec and personal firewall software. This chapter has not directly explored the trade-off between the different security approaches, but the next chapter does.
5. Static addressing is not necessary. It adds needless complexity with very little benefit in return. Network administrators must manage address allocation, and get directly involved in adding new systems to the wireless LAN.
   1. Static addressing provides only a minimal direct security benefit. Source IP addresses are not authenticated by the sender, and attackers are likely to learn the IP addresses being used on the wireless network unless you employ strong link-layer protection.
   2. Tracking users is better done through the user-based networking that 802.1X provides. With a username available to the network through the RADIUS server, there is no need to associate a user with an IP address. The user can be associated with the username instead.
   3. Dynamic addressing minimizes the chance that two users may accidentally be assigned the same address. Only one DHCP server is needed for several VLANs with judicious use of DHCP helper; if a DHCP server already exists, there may not be any reason to use another one.

Table 21-4 summarizes the different factors discussed in this chapter. Security is too complex to be reduced to a simple table entry, so it receives the full attention of the next chapter. As you consider this table and a purchase decision, keep in mind that some products work with certain topologies better than others.

[a] Newer products may reduce the backbone impact by logically attaching access points to a control device in the network.

[b] Both the single subnet and central concentrator architectures are typically used with VPN software for additional security. Obviously, if VPN software is used, the amount of client integration work is much larger.

[b] Both the single subnet and central concentrator architectures are typically used with VPN software for additional security. Obviously, if VPN software is used, the amount of client integration work is much larger.