September 10, 2009, 9:51 PM — The IEEE will shortly ratify 802.11n – a standard in development for six years and shipping in enterprise products for more than two. Much of the attention has been on the higher data rate, an increase from the 54 Mbps of legacy 802.11a/g to 300 Mbps, but the standard’s true implications are much broader – it signifies the advent of the all-wireless enterprise.
The changes in 802.11n are more radical than in previous upgrades, affecting every layer of the wireless stack. By increasing raw bandwidth through multiple antennas, wider channels, and lower protocol overhead, 802.11n can deliver a tenfold speed increase capable of handling all enterprise applications and devices. Early adopters have seen application-layer throughput of around 200 Mbps – twice as fast as Fast Ethernet.
But displacing Ethernet is about more than just speed and requires more than simply installing faster access points. Wireless networks also need the reliability, predictability, security and simplicity of wires.
To achieve this, wireless networks must be designed with two principles in mind: first, address interference at the RF layer; then give the network control of the client.
The need for RF-level interference control stems from multipath, the most far-reaching change in 802.11n. With 802.11a/b/g, it was relatively easy to estimate the degradation of data rates over distance. With 802.11n, signal strength depends on the way that radio waves reflect off, refract through or diffract around obstacles, which can result in unpredictable coverage areas that are spiky and discontinuous.
Unpredictable coverage areas make planning much more difficult than with previous 802.11 standards. Knitting them into a seamless pattern can be difficult if adjacent access points must use non-overlapping channels, a common requirement of legacy wireless LAN architectures.
Since you can’t design around what you can’t predict, radio management tools used in legacy wireless LANs don’t work well with 802.11n. Rather than hope that radio management software will avoid co-channel interference, the network itself must solve the problem at the RF Layer.
The second principle - let the network control the client behavior - results from growing client diversity. Whereas previous wireless standards offered relatively few client-side options, 802.11n supports multiple wave bands, numbers of antennas, numbers of spatial streams, channel sizes and many other settings. Tuning this bewildering array of options for optimal performance can be a major challenge in itself. And the difference between a well-tuned and a poorly tuned client is much larger in 802.11n than with legacy technologies.
In a shared medium, an ill-behaved device hurts not only itself but other devices around it. To ensure high and consistent throughput, the network must take control of client behavior. Only the network knows which access point is best suited to serve each client and only the network can adapt contention parameters to ensure predictability in a fundamentally unpredictable RF environment.
My experience over the past two years has been that an 802.11n network designed with these two principles delivers switch-like experience with the added benefits of mobility, allowing enterprises to migrate to the all-wireless edge. Leading enterprises across a spectrum of industries are already making this transition now. The rest will follow.
About the author:Dr. Vaduvur Bharghavan is founder and CTO of Meru Networks.
