802.11n 100Mbps WLANs: Some Day

As if there isn't enough confusion out there concerning the alphabetsoup of wireless standards, the last thing we need is the VP ofBroadcom's enterprise computing division talking about products based onstandards that are about as mature as an eighth-grader. A Broadcomspokesperson was subsequently quoted as stating the company had "noplans to preemptively launch products that could possibly undermine thestrong standards-based industry that we've helped build."

So what's really going on here? The 802.11n standard, which calls forWLANs with performance of at least 100 Mbps, is in its very early stagesof development and, given the immense complexities, I'll be shocked ifit is real before 2006. The technical hurdles are immense. Unlikeprevious 802.11 standards, which have defined the system's physicallayer data rate, 11n defines performance as effective throughput. Whenyou account for overhead from interframe spacing, packet preambles, MACheaders, contention and acknowledgments, today's 802.11 products havetypically achieved TCP throughput of about half the data rate (5.5 Mbpsfor 11b and about 27 Mbps for 11a). Further, this roughly 4x improvementin throughput must be achieved without decreasing effective range.Applying Shannon's capacity limit theory using today's modulationsystems, the maximum possible data rate using existing 802.11 channelsis 133 Mbps. Even if the underlying MAC design were radically altered,it would appear insufficient to achieve the stated goal of 100 Mbpsthroughput.

One solution to this problem is to use multiple RF channels, therebydoubling the available bandwidth. Both Broadcom and Atheros have usedthis channel-bonding technique in proprietary implementations to boostthe performance of their systems. But it is far from ideal: Throughputtypically increases by less than 30 percent; it makes dense APdeployment much more difficult; it reduces effective transmission range;and it makes legacy system coexistence much more difficult. Overall,it's a poor solution.

Most experts I have spoken with agree that newer, more spectrallyefficient wireless modulation technologies are far more preferable thanusing larger channels. It's highly likely that some form of MIMO(multiple-input multiple-output) technology will form the basis forfuture 11n standards. But today's MIMO implementations, while promising,fail to meet the 100 Mbps throughput standard, and they also can beexpected to suffer from other deficiencies normally associated withfirst-generation implementations.

I've long argued that network chip developers should be given somelatitude to add new features outside the definition of existingstandards as long as they maintain full compatibility with existingstandards and are willing to make these enhancements publicly availableas part of future enhancements to the standards. This approachencourages innovation and competition between chip vendors whiledelivering appreciable benefits for organizations that are willing tostandardize on a single product. As such, Broadcom should be applaudedfor its efforts to boost the performance of the company's wireless chipofferings, regardless of the technology it chooses to utilize. But toclaim that it will be the first to market with products based on astandard that doesn't yet exist, that's crossing the line.

-- Dave Molta, [email protected]