Ten years ago, designing a wireless LAN was more art than science. Most of the early wireless LAN deployments were “hand-crafted” in the sense that RF engineers designed the layout and placement of access points (APs) with simple tools and their good instincts.
Back then, when 802.11 networks operated at 1 and 2 Mbps and APs were “fat,” the rule of thumb for coverage planning in WLANs was 20% overlap of adjacent cells. That would result in a continuous coverage area, but the access points were far enough apart that they would not directly interfere with each other. Client roaming behavior was smooth and predictable. As wireless clients moved away from their access point and into these overlapping areas, they were able to detect that they were in the fringe coverage and would scan for better access points. In a well-designed network, a roaming 802.11 client would find a neighboring AP with better RF signal and automatically associate to the new AP.
Over the last few years, WLAN systems for the enterprise have become much more sophisticated. APs are now “thin,” and a central WLAN controller consolidates management, security and deployment functions. Modern APs are capable of automatically setting their channel, and have detailed control of their transmit power levels, even at deployment. Tools are available for 802.11a/b/g to plan networks, predict wireless coverage and recommend AP placement. The central controller can take a system-wide view and make adjustments as necessary to assure adequate coverage throughout the network. There is still some debate about whether this automated approach results in a better wireless network design than the good old handcrafted approach. However, the WLAN controller architecture has clearly driven the success of WLANs in the enterprise by reducing complexity and lowering deployment and operating costs.
So – what happens when 802.11n comes to the enterprise? Will we deploy WLANs the same way? Not exactly.
802.11n for the enterprise changes everything. There are many different facets of this change. New 802.11n APs capable of delivering ten times the capacity of the current generation will create tremendous load on enterprise networks and force redesigns for both wired and wireless networks. The debate about WLAN architectures – controllers and thin APs versus standalone fat APs – has resurfaced as WLAN infrastructure vendors migrate to 802.11n and optimize their systems to take advantage of the new capabilities.
Let’s focus on the range and coverage capabilities of 802.11n to illustrate how a new 802.11n deployment will be different from current 802.11a/b/g network designs. We previously covered the outstanding range of draft 802.11n products coming to market today for consumers and small businesses. They really do deliver at least twice the range of 802.11g products, and enterprise APs and clients based on 802.11n could do even better. But combining tens or hundreds of new 802.11n APs in an office may have unpredictable results in terms of the coverage achieved and the behavior of wireless applications.
There will be a wide variance of capabilities in the client devices using 802.11n. Certainly, legacy 802.11a/b/g clients will have very different range capabilities than 802.11n clients when communicating with new 11n APs. Most enterprise WLAN deployments will plan to support both legacy and new 11n clients for the next few years. However, even a greenfield 11n deployment could have a variety of client devices with differing range. There will be a significant range difference between an 11n dual-mode Wi-Fi/cellular handset and an enterprise 11n Wi-Fi client in a notebook PC. The notebook client will have multiple antennas built into the case and multiple radios operating at higher power. The dual-mode handset will likely only have one radio dedicated to Wi-Fi, and is not large enough to support the many antennas that are physically separated. The range of these products will be very different.
When designing an enterprise WLAN, how do you do coverage planning? Do you design your network for the least capable client? When 802.11n systems move into the enterprise, the disparity between the best-performing client and the worst-performing client will be greater than ever before, and new network designs must address this.
Environmental factors are more influential in coverage with 802.11n networks than with legacy 802.11 systems. The new physical layer for 11n employs MIMO (multiple in, multiple out) techniques to improve the link margin, and actually takes advantage of multi-path to improve the signal. Multi-path is essentially multiple reflections of the RF signal arriving at the receiver at slightly different times. Indoors, multi-path is caused by the radio signal bouncing off walls and other obstructions. 11n benefits from multi-path.
However, multi-path is a signal impairment for earlier versions of 802.11 – it’s essentially interference. A difficult location that was a “dead spot” or “coverage hole” in an 11g network may be a great location with excellent signal strength in an 11n network. Our intuition about designing network coverage for 11a/b/g networks is all wrong when it comes to designing 11n networks. The current set of automated network planning and site survey tools for 11a/b/g will likely produce strange results. The size and shape of the coverage areas with this new infrastructure will be very different. The coverage achieved by 802.11n will be more sensitive to building design, and will vary more from one installation to another. We will need to develop new tools and best practices for installing large enterprise WLANs based on 11n.
The method for achieving extreme range in 802.11n changes the behavior at the system level. With all of the optional capabilities enabled, an 11n client may be communicating to its AP with multiple spatial streams using multiple antennas and beam forming. Beam forming concentrates the RF energy in a particular direction, toward the intended recipient. The beam is formed dynamically and can be pointed automatically, creating the effect of a temporary high gain antenna between the two communicating stations. Range extension in this manner will change the shape of the effective coverage area even more. Coverage will literally follow a mobile client with smart antennas beyond the normal coverage area.
AP-to-AP roaming behavior will be different for new 11n clients with smart antenna systems. In current WLANs, the mobile client is usually associated to the closest access point. That may not be the norm for 11n networks. As a result, applications may perform differently. Overall, interference from adjacent APs operating in this manner may occur less frequently than with current systems, but when an enhanced mobile client is nearby and active, it will create interference for adjacent APs at a higher power level. Even interference in the enterprise will be different with 11n.
These are not huge problems. They just make the system behavior of WLANs in the enterprise very different than it has been for the last ten years. It is possible to deploy an 802.11n system in the enterprise without understanding or addressing any of these issues — existing 802.11 networks could upgrade their APs from 11g to 11n, placing all of the new APs in exactly the same locations. Legacy Wi-Fi devices would continue to work, though there may be funny behavior with some applications. Old trouble areas for coverage may be improved, and new coverage holes may emerge. New 11n Wi-Fi clients would be able to use the same infrastructure and operate at higher data rates.
Since new 11n APs will be more expensive than current thin APs, the simple AP swap-out could be an expensive system upgrade for an incremental performance improvement. It may be possible to achieve the same performance with fewer APs, or to have much better performance with the same investment in a system tuned to address the 802.11n system behavior.
To fully exploit the promise of 802.11n in the enterprise, we need to develop new tools, new intuition about range and coverage, and new best practices for WLAN design and deployment.
The technical part of the IEEE 802.11n standard is complete, though the official standard won’t be published for more than a year. Wireless chip vendors are already shipping their second generation 11n chipsets. Now, further innovation will come at the system level and there will be many new and different 11n enterprise WLAN systems introduced in the next few months.
It will be interesting to see how WLAN infrastructure vendors deal with the complexity, challenges and new capabilities of 11n. Fat APs may come back in style. WLAN architectures and deployment strategies that were once considered strange may become the new best practices for building WLANs.
N is for Novarum. Stay tuned for future installments in our wireless series as we examine the promise and pitfalls of 11n in more detail.