The End of Ethernet? Wrong QuestionThe End of Ethernet? Wrong Question

The discussion should start with a vision of the mix of service configurations we plan to support; along the way the role of wired & wireless LANs will work themselves out.

For some years we have batted around the idea that with advances in wireless LAN technology we could do away with the vast majority of our wired infrastructure and connect devices via Wi-Fi. John Cox has been hosting a lively discussion on Network World following a piece he did titled "Is it time to cut the Ethernet access cable?". This argument has gotten new impetus with the adoption of the much higher capacity 802.11n WLAN radio link that can theoretically push the raw data rate to 600 Mbps.The basic limitations of a radio system are capacity, reliability, and if the user is mobile, battery life. With earlier Wi-Fi implementations we could add security to that list, but the newer security mechanisms like 802.11i/WPA2 and 802.1x have addressed those concerns. The core issues are capacity and reliability.

In terms of capacity, a single fiber optic cable has more capacity than the entire radio spectrum, and the entire cellular telephone industry operates on less raw bandwidth than we have on a single Category 6a cable pair. The 802.11n standard defines a line rate of 600 Mbps, though today it's actually 450 Mbps in a 40 MHz channel as the current products have implemented a maximum of 3 rather than 4 transmit chains. In any event, we're still talking about a half duplex, shared media transmission system where the overhead reduces the maximum throughput to about 50% of the line rate. So best case, that half duplex 450 Mbps Wi-Fi link is going to give you roughly the same capacity as one full-duplex 100 Mbps wired connection, but now you're sharing it with everyone else who's associated with that access point. For most garden variety data applications, that's still pretty good.

Reliability is still tough to achieve, as sending electromagnetic radiation through free space is a dicey proposition, so radios will probably never be as reliable as wires or cables. However users seem to factor that into their thinking, and 802.11n's use of MIMO, MRC, and beamforming technologies will improve the reliability considerably over earlier Wi-Fi implementations.

Today, we are seeing a major move toward all-wireless in three primary environments: universities, health care, and hospitality. Universities are finding that Ethernet jacks in dorm rooms are on the same trajectory as wired telephones. A number of universities have removed the wired phones from the dorms and the Ethernet jack may be next. However, virtually all of the administrative workstations are still wire connected.

Health care is a special case given the mobility of the work force and the move towards more mobile health care systems. Health care is also the key vertical for voice over Wi-Fi, and hospitals are increasingly providing guest Wi-Fi access, which is securely partitioned from the internal users and applications. Again, the administrative workstations are still wire connected.

In hotels, you might still find a wired Ethernet connection in your room, but like the university environment, most customers ignore it and jump on the wireless connection. If I find a wired connection, I'm plugging in.

It is important to recognize these examples as special cases however, and not jump to the conclusion that this would also be a reasonable solution for a traditional office environment. Many large companies use the all-wireless approach for data access in small, rented offices where installing Cat 5e or 6 would essentially be a waste of money, and a shared 54 Mbps channel is all the capacity they really need.

For the large office environment, you really have to assess the options in a businesslike fashion, and voice is a big part of that consideration. Clearly there's money to be saved in getting rid of wired connections, and the biggest savings would come in a greenfield installation where we could simply reduce the number of cable drops and switch ports we would have to purchase. The big part of the equation everyone seems to miss is: where do we plug in the IP phone?

Unless we take the big step of assuming that all voice traffic will be carried on a wireless network (or that people will simply give up talking), we're still going to need wired ports (probably with Power over Ethernet capability) for phones. The future of the desk phone is one of our big question marks in the migration to UC, but even if it's a softphone, it will need a reliable connection and enough transmission capacity to do voice and potentially video as well. In a WLAN, the softphone gets an additional boost because it can actually use the PC's 802.11n connection; given the power requirements, we don't anticipate having 802.11n voice handsets for a few years at least.

Fortunately we do have handoff and quality of service mechanisms for Wi-Fi (i.e. 802.11e/WMM), but capacity will still be an issue, particularly if we add video to the mix. Frankly, I cannot see a laptop being used as a mobile voice device, so the combination would most likely be a PC or laptop and an ancillary mobile device. That mobile device might work on cellular, Wi-Fi, or both. We do have a number of strategies for merging the fixed and mobile environments, but each involves issues of cost and trade-offs in terms of functionality

Conclusion The bottom line is that when we are talking about whether we can eliminate wired connections, we're asking the wrong question. What we really need to do is look at how we intend to provide mobility and integrate it with our overall plans for delivering the full range of unified communications services. That discussion should start with a vision of the service or more likely the mix of service configurations we plan to support, and along the way the role of the wired and wireless LANs will work themselves out.The discussion should start with a vision of the mix of service configurations we plan to support; along the way the role of wired & wireless LANs will work themselves out.