The Crippling Effect of Mixed Mode
April 05, 2004
Tests show that running a variety of products mixing 802.11b and 11g on a single channel could be severely restricting overall wireless network throughput.
It has long been acknowledged that mixed-mode wireless networks -- those that simultaneously run clients with both 802.11b and the five times faster 802.11g, which both operate in the 2.4GHz radio band -- can suffer speedwise. Specifically, the use of the slower 802.11b can drag down throughput on 11g.
In a series of tests conducted with off-the-shelf enterprise class access points and up to four concurrently running client systems in various configurations -- everything from a full-speed 11g notebook down to a PDA running 802.11b at only 1 Megabit per second (Mbps) -- revealed that just one client running at an extremely reduced data rate can bring all the clients down significantly.
802.11g clients are rated to run at 54Mbps, though they usually get only around 20 to 25 Mbps in real world use. 802.11b clients are likewise rated for 11Mbps, but top out at around 6Mbps. The farther away the devices are from an AP, the lower the throughput drops, as well.
Kevin Tolly, the president and CEO of Tolly Group, says he expected significant results from his experiment, "but they were more significant than we anticipated."
He says that any wireless devices running at a higher speed is going to be impacted by a slower device on the same channel.
"We think of radio frequency (RF) as being out in the open and not constrained -- but it is," says Tolly. "It's like the Lincoln Tunnel. I can walk through the tunnel at one mile per hour, or you can drive at 50. But if I'm walking ahead of you, you have to wait." The faster device is still capable of the higher speed, but is stuck in a queue behind the slower device.
Testing was done with both TCP and UDP protocols, and UDP scored better. However, TCP is the underpinning of most real world network traffic these days.
Scott Lindsay, the vice president of marketing at Engim, says that since wireless LANs are a shared medium you have to squeeze out the throughput where you can, since there's "not much spectrum to go around. Some IT managers I've spoken with, they didn't even know it was shared. They're used to wired, where you put in 10/100 cable and that's what you get."
Engim, of course, wants the word to spread on these findings as its chip design is meant to counter these effects by using up to three channels at once fro a single AP. Still favoring traffic metaphors, Lindsay describes Engim's solution as the equivalent of a three-lane highway handling traffic, compared to the one lane of traffic found on most APs today, including those tested.
Tolly's test results (available online) show an Engim-based AP getting as high as 50 times better throughput.
Engim's chip is in production, but won't appear in any products until at least the third quarter of this year.
Tolly says that Engim's solution solves the "wireless weakest link theory" but notes that other's copying the three-channel technology could lead to patent infringement -- though the US Patent and Trademark Office Web site shows only one current Engim patent, and its not for three-channel simultaneous use. No matter, as the expected solution is for Engim to license the ability to chip vendors in the future.