From convergence to mobility to virtualisation, every networking innovation that delivers convenience to end users also brings complexity for IT managers. Nowhere is this tradeoff in complexity greater than in the migration to IEEE 802.11n wireless networks.

Though wireless networks have always seemed unpredictable, the effect is magnified by the many combinations of client-side options that 802.11n permits – more than 5,000 in Wi-Fi certified devices so far, a number that is growing all the time.

Each of these 5,000 client configurations has its own unique behavior, all responding to changing radio, contention and bandwidth conditions in different ways. Coupled with variable coverage and loss patterns in the unlicensed bands, tuning client configurations can make wireless networks seem impossible to manage.

Yet the demand for reliability in wireless networks has never been stronger. With 802.11n, wireless is becoming a primary strategic network from which users demand guaranteed availability and assured service. Client troubleshooting has consequently become the single biggest problem for IT administrators. For IT to deliver wireless service like wired, the dark arts of radio frequency management need to become a science.

In terms of operational and troubleshooting complexity, wireless management tools have hitherto failed to keep up with either the wireless LAN technology itself or user demands for utility connectivity. Most tools still rely on reactive troubleshooting: a user reports a problem, then a technician tries to recreate it and identify the conditions that led up to it – something that’s rarely fast or simple in a constantly fluctuating radio environment. The hours and days spent trying to find and fix wireless network faults don’t just represent downtime and lost productivity, they entail thousands of dollars in ongoing expenses, costs that grow with every new client, access point and application.

To curtail costs and ensure reliability, three things need to change.

  • First, management needs to become proactive to deliver service assurance to wireless users: Network users and support staff need a way to be sure that a wireless network is meeting their applications’ performance criteria, spotting potential problems before users experience them.
  • Second, the network needs to keep track of detailed RF history so that if a problem does occur, its causes can be traced quickly.
  • Third, as networks scale, it is impossible to rely on human intuition to resolve problems. Instead, the network must mine its own data and surface issues – rather than relying on human intervention. Insight with tools is mandatory to scale.

Service assurance is relatively easy in wired Ethernet. Because the network path is well defined and endpoints are static, IT can run end to end tests to make sure the pipes are clean before users arrive. It is much harder in wireless networks, where predictability often extends only from the controller to the access point. Beyond that, the air link is continuously changing and the client device may be unknown. The two ways to be sure of wireless performance have been to either send people out to walk around with laptops and inject traffic, or use overlay sensors to play the role of client devices. The former approach is prohibitively expensive, while the latter approach requires an independent deployment with attendant costs and complexities of its own. Neither approach is scalable, and the onus falls on the wireless network itself to detect its problems proactively.

The next generation of wireless infrastructure must use access points themselves to inject traffic. In this approach, each access point runs a virtual client that connects to its neighbors, sending real traffic over the same airwaves and backend security infrastructure that real clients use. Unlike manual testing, a virtual access point can run 24/7, gauging the exact performance of the network as loads vary throughout the day or week. And because it uses access points themselves, it scales as the network grows.

When a problem occurs, IT staff need to be able to “rewind” the network state to understand its cause. This requires a database of all significant network and RF events, essentially integrating wireless sniffer-like capabilities into the infrastructure, with a scalable store and retrieve mechanism.

In order to achieve all these features, the RF network needs an architecture where neighboring access points are on the same channel and use proximity to both talk to and snoop on each other. This will allow applications to run correctly, assuring users that service is available. Just as a dial tone lets people know that the phone system is working, proactive wireless management will give users and IT staff confidence that 802.11n networks are up to speed.