Sep 11, 2007 Does 802.11n spell the 'end of Ethernet'?
Does 802.11n spell the 'end of Ethernet'?
Analyst says 802.11n is 'good enough' for wireless client access
Is the advent of the 802.11n wireless standard the “end of Ethernet” - at least in terms of client access to the LAN?
That’s the provocative title, and thesis, of a new report from Burton Group, written by senior analyst Paul DeBeasi. He began looking into the question when he heard a growing number of clients asking whether it was time to discontinue wired LAN deployments for connecting clients. Would 11n, the next generation high-throughput Wi-Fi, make the RJ45 connector in the office wall as obsolete as gaslights?
On the surface, the question seems to answer itself. Enterprises almost as a matter of course have been focused on making the wired connection to clients as capacious as possible, at as much as 1Gbps Ethernet. The first WLAN products based on draft 2 of the 802.11n standard are demonstrating throughput of 150Mbps to 180Mbps, which has to be shared by the number of clients connected to a given access point.
In addition, DeBeasi’s report also details how WLANs introduce some novel problems, such as RF management and new security risk vectors, and complicate existing ones, such as network management and IT training. All of which makes one ask, “Why Fi?”
DeBeasi has a simple answer. “The mobility piece,” he says. “I have a 21-year-old, and an 18-year-old, and they have never plugged [a computer] into anything in their lives. And I have wired Ethernet in my home,” DeBeasi says. “They all expect ubiquitous Wi-Fi. That’s the unstoppable force that’s pushing us forward.”
“Mobility is already a priority for my organization,” says Jeffrey Allred, manager, network services, at Duke Comprehensive Cancer Center, part of Duke University Medical Center, Durham, N.C. “MDs, PhDs, and researchers in general are all over the place most of the time and being ‘connected’ from wherever they may be is critical.”
It’s also a priority at Grant Thornton LLP, Chicago, the U.S. member firm of the global accounting, tax and business advisory firm. “As a professional services organization, it is important for our people to collaborate at anytime and anywhere,” says Michael Ruman, IT messaging manager for the firm. “Adding wireless to the corporate infrastructure has allowed our professionals to move from one physical office to another and conduct meetings or work sessions… anywhere within the organization.”
The Burton report tackles the performance differences between Gigabit Ethernet and 11n head-on.
“However, does greater throughput really matter to most enterprise users who are not transferring enormously large files?” the report asks. “We found that although 802.11n was slower than Gigabit Ethernet, the download time difference was negligible. Even with 20 users per [access point], the file download times ranged from two to eight seconds -- still satisfactory for most users… Burton Group expects that 802.11n will eventually approach the performance levels provided by Fast Ethernet and that, for many enterprises, 802.11n throughput will be good enough.”
Does DeBeasi think “good enough” is, well, good enough for enterprise nets? “If you look at 11n with 150Mbps and 20 users sharing the access point, they get 7Mbps average throughput,” he says. “They don’t get that in their homes with DSL and cable modems. It’s time for people to reset their thinking.” As part of that reset, DeBeasi points to the report’s comparison in VoIP performance between Gigabit Ethernet and 11n in latency, which is the one-way delay between a sender and receiver on a network, and in jitter, which is the amount of variation in the arrival times of VoIP packets. In both cases, he acknowledges, 11n is considerably worse than Gigabit Ethernet: latency is roughly 20 times that of Gigabit Ethernet, and jitter can be as high as 150 times.
“But who cares?” asks the report. In both cases, the absolute value of the 11n results is still only a small fraction of the wireless VoIP “budget.” Both latency and jitter in 11n should be “good enough,” DeBeasi argues.