Week of June 28, 2006  

Mobile Observer
June 28, 2006
Wireless Propagator: A Wireless Repeater to Switch Transition? -- Part 1
By Frank Bulk (mailto:fbulk@nwc.com)
Blurb: As wireless networks continue to spread and become more pervasive in 
organizations today, the wireless network manager needs to understand the 
caveats of a micro-cell infrastructure and some other options.
Perhaps it hasn't quite happened at your workplace yet, but some organizations 
are shifting the balance of their network traffic from the wired to the wireless 
one. Telecommuters or road warriors who visit the office once or twice a week 
(or month) can now work out of a temporary cubicle in the same way they do in 
their home offices, on the road or at a hotel: Turn on their wireless card and 
associate to the network. 
In other locations, Wi-Fi is enabling applications and services that weren't 
available or affordable before. Hospitals are able to take advantage of VoWLAN 
phones from the likes of SpectraLink and Vocera to enable instant and directed 
communication to provide better bedside assistance. Universities are making 
Wi-Fi available in the classroom and public areas so students can have quick and 
convenient access to information or engage in class assignments. 
Basically, as usage increases, the wireless network needs to morph from one 
designed around coverage to one designed around capacity, from spare to dense.
It's been almost two years since Aruba began heavily marketing the "wireless 
grid," a term for its micro-cell concept that emphasizes smaller and overlapping 
cells of wireless coverage. The other leading enterprise WLAN vendors recommend 
roughly the same design for dense installations. Rather than leave APs (access 
points) on their default output power of 100 mW, the APs (automatically) can 
turn down their power so the wireless signals don't propagate as far. At the 
same time, the vendors tout their AP placement planning software, auto-RF tuning 
and load balancing, all while selling a few more APs and supporting wireless 
controllers along the way. 
The micro-cell concept is conceptually intuitive, but the devil is in the 
details. If a three-channel design (1, 6 and 11) was complex the first time 
around, designing a dense wireless network requires many more access points, 
which significantly increases the complexity of AP placement. More importantly, 
the potential for co-channel interference rises, for at least two reasons. 
First, even if the access points are configured to link at 36 Mbps, to establish 
a minimum bound for performance reasons, the APs' basic rates for management 
frames can be as low as 1 or 6 Mbps, which transmit decipherable information 
much farther than the higher rates. Compounding the problem, client cards are 
usually able to pick up RF energy up to twice as far as the limits of the 1 Mbps 
range. This is enough to fail the client card's clear channel check, which in 
turn delays transmission. 
Second, unless clients support automatic client output power control -- as is 
the case with CCX (Cisco Compatible Extensions)-enabled cards and Cisco APs -- 
they will not automatically adjust their output power, which leads to the 
problem of power asymmetry. The client's Wi-Fi card, operating at full power, 
essentially broadcasts far beyond the range of the lower power access point to 
which it is associated. This means a neighboring AP on the same channel will 
hear it, too, as will many of the other clients on that channel associated to 
the neighboring AP(s). The IEEE is working on standards that will enable 
AP-client link management communications, but we're stuck with status quo for at 
least the next two years.
These co-channel problems are not new. I had personally hoped that enterprises 
would require 802.11a radios in their second-generation wireless networks and 
take advantage of the dual-radio client cards embedded in many laptops. However, 
although the majority of enterprise APs are dual-band, many are not dual-radio, 
which means it's an either/or instead of a both. Intel will very likely pump out 
only dual-band Centrino chipsets moving forward, but it seems that most of the 
time the drivers are not configured by default to prefer 802.11a. 
Unfortunately, our most recent surveys confirm that not many organizations, not 
to mention end users, are taking advantage of 802.11a. It's a pity, because most 
implementations of 802.11a support at least eight nonoverlapping channels. This 
significantly simplifies the channel selection process and reduces co-channel 
interference, and the higher attenuation characteristics of that frequency are 
conducive to a micro-cell design. Hopefully, 802.11n, which can take advantage 
of both the 2.4- and 5-GHz bands, will obfuscate the spectrum issue and make 
channel choices a moot point. Think of it as you would the three bands our GSM 
cellular phones use, which we don't worry about.
Another possible solution is to use a four-channel design, an option that Aruba 
recently re-introduced. In a lecture hall with 250 students each using 
school-provided laptops, it's understandable for an enterprise wireless 
infrastructure vendor to want to reduce the per-AP load below 80 clients. 
According to Vice President of Product Marketing Keerti Malkoote, Aruba has 
tweaked its product to minimize the adjacent-channel defects generated by the 
slight channel overlap found in the 1, 4, 8 and 11 line-up. It sounds to me like 
a semi-desperate attempt to deal with density issues, and it's not to be 
recommended. (For some primary reading on this topic, check out: 
http://dades.joor.net/article_fitxers/1253/FourPoint.pdf and 
http://www.cisco.com/univercd/cc/td/doc/product/wireless/airo1200/accsspts/techr
ef/channel.pdf )
Two vendors have followed the coordinated single-channel route: Meru and 
Extricom. Both use coordinated access of the channel between APs to scale the 
number of clients and provide predictable levels of service, essentially 
eliminating the penalties of roaming between access points. And each scales in 
capacity by overlaying the network with another channel. Competing vendors have 
thrown some fierce barbs over the years because of the companies' admittedly 
unique and unconventional technology, but customer installations thus far have 
substantiated Meru's claims and Extricom's installed base is decidedly smaller.
Adjusting a network from casual and light use to mission-critical and pervasive 
use will mean some growing pains. In three weeks we'll talk about how you might 
possibly convert those wireless APs from hubs to switches and how new antenna 
technologies will help reduce the problems of general and co-channel 
interference.
Frank Bulk is a contributing editor to Network Computing Magazine covering 
wireless and mobile technologies and works for a telecommunications company 
based in the Midwest.
For more analysis and opinion from Frank Bulk.
http://www.networkcomputing.com/channels/wireless/