Enterprise Wi-Fi Architecture--Proprietary from Edge to Core

Remember the battle between thin and thick access points? It's now been almost five years since Symbol Technologies, once the leading enterprise WLAN innovator, launched its Mobius "wireless switch," a diabolically deceptive name if there ever was one. Startup vendors Airesapce, Aruba Networks, and Trapeze Networks quickly followed suit, with similar product offerings packing enhanced functionality. The battle for best WLAN architecture continues to this day, and with new technologies like 802.11n and dual-mode telephony on the long-term strategy agenda of most enterprise network planners, the war continues. It's largely a battle for customer lock-in, because when it comes to enterprise WLANs, switching packets is a whole lot easier than switching vendors.
At first glance, the evolution of enterprise WLAN product architecture appears to mirror the evolution of enterprise Ethernet. Those of you with grey hair are likely veterans of the early days of Ethernet, when the physical-layer transceiver was distinct from the media access control (MAC)-layer network interface. Increased market competition spurred chip design innovations, and before too long, both the PHY and MAC were integrated. This made for easier and more reliable network designs and spurred the market for 10BaseT and, later, for what was arguably the most significant networking invention ever conceived: the Ethernet switch. More recently, commoditized Ethernet switch functionality has moved further up the stack, providing many new challenges and opportunities. Along the way, mainstream Ethernet performance increased 1000-fold or more and the basic architecture of enterprise Ethernet systems has been broadly embraced across the industry.

Enterprise WLAN architecture evolution shows some parallels but also some fundamental differences. While the pace of performance improvement has lagged that of Ethernet, speed increases have nonetheless been impressive. First-generation products delivered throughput of a little over 1 Mbps. New 11n offerings will offer 100 times that number. However, while Ethernet made the transition from shared to switched medium, Wi-Fi continues as a shared medium technology, requiring complex coordination of physical and datalink protocol behavior throughout the network. Even worse, the medium is hostile, subject to interference from other WLANs as well as myriad RF-emitting devices.

While the IEEE continues its 10-year battle to correct deficiencies of the original 802.11 standard, vendors have also found themselves waging a never-ending battle on multiple fronts, trying to spur increased enterprise adoption. First, the thin-AP was born, initially trumpeted as a less expensive solution and then, when that didn't turn out to be true, as a more manageable architecture. However, individual vendors all made their own decisions where to slice the protocol stack between AP and controller, resulting in numerous design deficiencies and interoperability obstacles. Yes, there are challenges associated with deploying multi-vendor Ethernet networks, but many organizations do it, either because of happenstance (e.g., a merger) or design (to optimize performance or keep vendors honest). With enterprise Wi-Fi, multi-vendor networking is pretty much impossible. Enterprise Wi-Fi spells vendor lock-in. That favors a few major players and discourages competition from startups.

Yes, there are industry initiatives designed to promote interoperability. The Wi-Fi Alliance's product certification program is one notable example, but while it is useful, especially in the consumer market, it only certifies basic protocol interoperability. It doesn't certify full standards compliance and it doesn't come close to certifying enterprise product interoperability. A more interesting example is the IETF, where the CAPWAP (control and provisioning of wireless access points) protocol is being hammered out by a working group. Early in the design of CAPWAP, and not for lack of effort from the many participants, it became clear that a meaningful industry consensus would not be reached. That's because the architectures of today's leading enterprise WLAN solutions are so fundamentally different from one another. If you compare the designs of Cisco Systems, Aruba, Symbol, Meru Networks and Trapeze, you'll find plenty of surface similarities, but far more differences under the hood. Even if the CAPWAP working group could define an incredibly flexible architecture for provisioning and controlling AP's, the probability of such a standard providing meaningful interoperability is quite small indeed.

In some ways, it's not such a bad thing to see the major vendors sniping at each other, calling each other out for product design flaws. In almost every case, architectural design decisions employed to solve one problem have negative consequences in other areas. And in most cases, the design innovations come at a very high cost. Vendors are still largely captive to the fundamental 802.11 MAC design and the laws of physics that govern RF. For enterprise Wi-Fi, there really hasn't been a breakthrough innovation that can compare to the Ethernet switch so, in the meantime, vendors are forced to cobble together complex, proprietary solutions.