John Cox has an interesting rundown on large installations picking 802.11n for client use instead of upgrading or adding more Ethernet: Cox, in a Network World story, starts with the observation that some companies and many colleges are finding huge numbers of unused Ethernet ports, which means they're paying depreciation and operating expenses on gear they're not using.
One school he speaks to has 80 to 90 percent of its Ethernet switch capacity unused. The CalState system performed careful analysis of current use and opted to cut 2,500 switches, which will save $30m in hardware-related spending, exclusive of HVAC/electricity savings.
At colleges, this is a simpler matter, because campuses can simply eliminate new spending for Ethernet in dorms and elsewhere and pull switch plates and switches, or reduce the number of jacks by a large number without impairing functions. While some students might have desktop computers, surveys keep finding that most arrive at campus with laptops or purchase them on arrival. Businesses may still need to have Ethernet active because of their heavier desktop use.
For instance, in a 2008 survey of incoming students at the University of Virginia, with 95 percent of students surveyed, a single person in the 3,071 asked did not own a computer. That number was as high as 26 percent (634 of 2,437) in 1997, but has dropped to a negligible amount since 2003 (30, 10, 18, 4, and 4 in successive years had no computer).
Now out of those 3,070 computer owners in 2008, only 36 had desktops. That's a lot of spending on Ethernet for 1 percent of students. And those desktop systems might have had Wi-Fi built in if they were Macs (Mac Pros are the only model that requires an add-on build-to-order Wi-Fi adapter) and most student/entry-level oriented Wintel systems.
802.11n found its way into colleges quite early, but enterprises now have a wide range of affordable options from major and minor vendors alike that are proving more cost effective than 802.11g or a/g was because of the greater capacity and range of 802.11n. Everything I hear from companies and read in reports shows that dual-band 802.11n overcomes almost ally of the gating factors that made 802.11g useful but not strictly a wired replacement for clients.
Most clients don't need anything like the 100 to 150 Mbps throughput that 802.11n can offer in ideal cases. Rather, each client may need from 1 to 10 Mbps in a more or less reliable and guaranteed fashion, and with a multi-channel switched WLAN, enterprises can easily deliver that.
College campuses have lower requirements, seemingly, with Cox noting that 1 Mbps is a reasonable threshold for common activities. In those cases, you need networks that can support massive concurrent users in relatively small areas, like classrooms or quadrangles, a very different requirement from the business network.
No one's suggesting Ethernet will be pulled out. It's still the only way to run critical services, and you need quite a lot of it to backhaul all the WLAN systems that are being put in. But there's a growing divide between client Ethernet and server/backhaul Ethernet that can let companies and colleges trim their IT budgets without reducing utility for their users.