Pest Control: 802.11n's Dirty Little Secret

There's a big elephant in the Wi-Fi room and it's not just Cisco. 

But no one (read vendors) seems to want to talk about it in front of customers.  We will. 

The biggest problem with 802.11n systems today is lack of consistency of performance. There's no getting around it.

Under certain conditions 802.11n devices yield tremendous performance gains over older .11g/.11a systems (and these are the numbers vendors typically quote), but under a wide variety of conditions that are typically encountered in real world deployments their performance suffers.

Inconsistent performance of .11n is one of the huge unspoken problems in the Wi-Fi industry. No one wants to talk about it since most vendors don’t have adequate means to address the problem. 

Most vendors – both from the systems and chip side – would like customers to believe this is all being solved by interoperability issues getting hammered out as we move into 2^nd and 3^rd generation .11n implementations. While that certainly has been a significant factor affecting performance consistency to date, as those issues slowly get resolved the remaining issue of dynamically managing the multipath environment to ensure predictable spatial streaming becomes the 800 pound gorilla sitting in the corner.

Have you seen today's 802.11n APs with all these di-pole, omni-directional antennas sticking up?  Each one can be "articulated" in a different direction. The problem is, how the hell is an IT person supposed to know which one to point where?  The answer is: they have NO idea (and they shouldn't!).

And even if it were possible for a normal IT guy to understand this, how the hell (we say hell a lot at Ruckus) would you manage a large deployment with hundreds of access points?

Any informed CIO or IT manager in the enterprise space that is considering an 802.11n deployment should question why the six antennas that protrude from the typical dual band .11n AP that you get from Cisco or other enterprise WLAN vendors includes very flexible ways to adjust the orientation of the antennas - yet there are no instructions on how these should actually be adjusted, under what conditions they should be adjusted, and what difference does it make in terms of the operation and performance of the AP.

Beyond that, this variation in performance is dependent on which client the AP is communicating with, how the client is oriented, whether there are any changes in physical environment occurring in real time, etc.

The net-net is that there is no possible way to statically configure the position of fixed antenna systems that will adequately ensure consistent .11n performance under real world conditions. This is why no equipment vendor wants to talk about this.

We’ve found that relatively minor changes in the orientation of the typical rubber duckie type dipole antennas that are provided on typical enterprise .11n access points can introduce MASSIVE variations in performance with respect to a given client.

Anyone can easily demonstrate this by just charting throughput as a function of antenna position or by rotating an AP to slightly different positions particularly at moderately difficult locations.

We benchmark this type of phenomenon all the time in our lab as a point of comparison with respect to how our products stack up against the market leaders, and the performance variation of competitive enterprise WLAN products as you vary the antenna and access point position is huge. But no one believes us, because we're the only people that have really solved this problem. Go figure.

But screw it.  Test one and see for yourself.  If you know what you're doing, you'll see we aren't full of poopage.

Comments

[Devin Akin]

This is so true...and funny as hell. Oops, I guess I say hell a lot too. ;)

[Jack]

I think the access points shown in the picture are very cool looking! Are they yours? Can I buy one?

[Oscar]

This article is spot-on. The touted 802.11n performance gains are largely due to MIMO technology. MIMO performance is significantly affected by several factors including antenna positioning, the frequency being used, location of the client PC(s) relative to the access point, and the multipath environment (how radio signals bounce off of the walls in your home). To the average user, finding the best location and antenna positioning for your wireless router is equivalent to adjusting the rabbit-ears on an old TV set (it's effectively random). While it is possible to develop DSP algorithms to dynamically adapt to various signal conditions, conventional algorithms are very complex and increase the cost of the electronics. There are some companies (Motia, Comware, etc.) working on more cost effective DSP technologies, but the bottleneck is in the standardization process and the surrounding politics between the major IC vendors. Unless you have an advanced degree in signal processing you may be better off sticking with your G router for another year or two until the MIMO experts come out with a more robust and cost-effective solution.

[Devin Akin]

Very good article. Inconsistent coverage is a disaster and few vendors address it.

One exception to your antenna orientation statement is Meru. In their deployment guide, they have a page dedicated to showing you how to deploy their antennas "orthogonally." However, they will tell you that the differential between orthogonal and all-vertical orientation is basically irrelevant...other than the fact that orthogonal looks WAY cooler. :) I'm not convinced it's irrelevant because EVERY time I see them use their own APs (demonstrations for example), they're deployed orthogonally.

Your point of testing it is well-said. We're on it.