02 August 2009

Who Moved My Packets, Or How I Learned To Stop Worrying And Cut The Cord


With the ratification of 802.11n just around the corner, it’s a good time to reexamine the fundamentals of Wi-Fi design and determine how this blazingly fast new technology will affect you. Who Moved My Packets is about the design considerations associated with 802.11n data, voice, and video applications.

Let’s start with a discussion about designing for coverage or capacity. For some wireless applications simple connectivity is the biggest issue with which users have to contend. Designing a network for coverage ensures that a Wi-Fi signal can be received at any location in which a Wi-Fi device is likely to be used. Connectivity is the primary objective - bit rate, packet throughout, multi-media support, quality of service, and even redundancy of coverage are secondary considerations.

Consider an indoor application in which Wi-Fi is used to communicate with a pool of bar code scanners for inventory management. The users are few in number, the amount of data transmitted is relatively small. Since the bit rate of an in-building Wi-Fi connection typically falls with distance and in the presence of interference sources, what started as a high speed connection near an access point could drop to 1Mpbs or less just a short distance away. However, even at that low throughput, a network designed for coverage should be sufficient for the application.

Any Wi-Fi network can be designed for coverage, and as a rule, designing for coverage requires far fewer access points. Just crank up the access point power to full, space the access points so that their coverage patterns overlap slightly, and the design part is done. Interference compensation, fair airtime availability, security, and network management are another matter entirely, but they’re outside the scope of this discussion.

A handful of Wi-Fi vendors have made an art of promoting their products as requiring fewer access points. Some Wi-Fi array (multiple access points in one box) and single channel vendors go so far as to tout their “unique” ability to deliver what no other Wi-Fi vendors can accomplish.


It’s all smoke and mirrors. Wi-Fi vendors all use Wi-Fi chip sets from a small pool of IC suppliers, and by regulation the power output of the radios is tightly controller by the government. The distance over which they can transmit, using comparable antennas, is the same. If you pull back the curtain, the secret of their claims is simply that they’re designing for coverage. Nothing more.


In fact, it’s really something less. Why? Because many users need a system that is designed for capacity. In a network designed for capacity, coverage is a given but bit rate, packet throughout, multi-media support, quality of service, and often fault-tolerance are primary considerations.


A capacity-based network requires that the vendor pay keen attention to internal architecture, algorithmic processing, and packet handling necessary to service deployments with a high capacity requirement: (1) large number of users; (2) users that are densely congregated; or (3) applications using voice or streaming video or business-critical telemetry data. Coverage alone is not sufficient for these scenarios – they require guaranteed bit rate, high packet throughout, and quality of service.


These scenarios are already the norm in education, healthcare, and government applications, and are fast becoming typical in enterprise, retail, and industrial deployments. With the migration of data, voice, and video applications to 802.11n from wired LANs, the need for capacity-based Wi-Fi will skyrocket. Users will expect wire-like performance with virtually unconstrained capacity on their shiny new 802.11n networks.


So the next time you’re given a pitch for a wireless LAN with one half, one quarter, one eighth the number of access points of an Aruba network, ask the vendor if they’re designing for coverage or capacity. And ask for test data to back it up. Doing so will avoid following Maj. T.J. 'King' Kong on a ride that is a mistake from the outset.