I recently had the opportunity to examine the coexistence opportunity enabled by .11n, to wit, the possibility of supporting a .11g client and a .11n client in the same channel on the same AP simultaneously. I've always thought this is a bad idea, largely based on experience with trying to do the same thing with .11g and .11b. Both clients are slowed to a very significant degree. Sure, it's possible to implement scheduling algorithms that implement fairness, however that's defined, but the question remains - how much degradation will we see if .11g and .11n try to operate in the same channel simultaneously?
To find out, I set up a simple testbed consisting of Cisco infrastructure and PCs using various Intel Wi-Fi clients. Operating just one .11g and one .11n simultaneously in a 20 MHz. channel cut aggregate throughput by more than two-thirds over that obtained from a single client operating alone. This is also significantly worse than the typical aggregate throughput we've seen in the past using two clients operating using the same technology. So my opinion is unchanged - don't try to mix technologies, and put your shiny new .11n clients on their own channel.
And that channel should really be in the 5 GHz. bands, since most installed WLANs today use the three non-overlapping Wi-Fi channels at 2.4 GHz. (and sometimes four slightly-overlapping channels; this can also work very well). Cisco has a clever feature in their controller called BandSelect, which will automatically move dual-band-capable clients to 5 GHz. with no client reconfiguration or user action of any form. But no matter how you get there, 5 GHz. should be your destination.
All of this is discussed our latest Farpoint Group Technical Note, dealing with practical .11n deployment strategies and a number of benchmark testing results. I wouldn't say this is the definitive work on the subject, but the evidence mounts - .11n is all you should buy now, and it should be deployed at 5 GHz. in most settings.
Mathias is a principal at Farpoint Group, a wireless advisory firm in Ashland, Mass.
The Leader in Network Knowledge
only one AP ?
Did I read correctly ? Only one AP for testing purposes ?
I shall say that the perf. downgrade may come from inside the AP, not from air perturbations.
There should have been a 11a+11g simultaneous test (for a better baseline/reference set) and also 11g and 11n activated on two separate AP (or better, 2 separate systems).
Possibly, but...
The point was to determine if .11g and .11n can coexist on the same channel simultaneously, meaning in close proximity, so a single AP is the only way to go for that purpose. To be complete, we really should try more simultaneous clients and more and different APs. While I don't expect the results to be all that different - stay tuned.
There would obviously be less of an issue using two different systems altogether - just mutual interference. But that's not a realistic scenario in an enterprise.
Thanks for the note.
Craig.
Post new comment