|
While many have assumed contemporary switched or controller-based implementations represent the end of the road with respect to WLAN architecture, the high degree of variability evident in modern controller-based architectures should serve as an indicator that such is hardly the case. Indeed, innovations in wireless LAN architecture continue to appear, as is witnessed by the announcement of a distinctive new approach from Wi-Fi startup Aerohive Networks. In some ways, the Aerohive model is reminiscent of the original traditional/robust access point (AP) approach - we refer to these architectures as distributed, in opposition to the controller-based centralized model that has dominated the thinking of the WLAN industry for the past seven years. In reality, both alternatives encompass a high degree of architectural variation, and we’ll explore the features and benefits of each in more detail below. Nonetheless, we believe the Aerohive architecture represents an interesting and innovative new direction in WLAN systems design, with a combination of elements that define an interesting and important approach to wireless-LAN system architecture.
The Evolution of Wi-Fi System Architecture
Early wireless LANs, before 1992, and with the exception of the first microcellular products from the Canadian firm Telesystems SLW, were almost always based on a peer-to-peer client model. A purely peer approach has major problems with scalability, in that all peer nodes in a given network must be able to “see”, in a radio sense, every other node. This severely limits the value of the peer model – even though it survives today as ad hoc mode in the IEEE 802.11 standard, and is a widely-recognized security hole when enabled. Access points were initially used primarily as bridges between wireless clients and a wired infrastructure, rarely with support for roaming. As roaming between APs became a core justification and requirement for the installation of WLANs, providing the ability to grow and scale a given installation smoothly and economically, APs themselves consequently became more complex.
Each had its own IP address, and each was configured independently. Soon a number of other issues with this “traditional” AP model became apparent, as follows:
- Security – Wi-Fi security, no matter what form, only covers the airlink, that is, the connection between wireless client and an AP. It leaves the remainder of the network value chain either unprotected or in the domain of higher-level security techniques, such as virtual private networks (VPNs).
- Management – Each AP is individually managed. While it is possible with some products to “make one AP look like another” and/or to use a central management tool or appliance, such is not the default in most cases, leading to serious management concerns once more than a few APs are deployed in any given case.
- Roaming – Most APs lack support for Mobile IP or a similar mechanism for provisioning mobility across subnet boundaries. Of those that do, configuration can become very complex.
In the early 2000s, however, most WLAN vendors realized that the high degree of common functionality in access points provided the motivation for moving much of this capability into what became known as a wireless switch. This in turn allowed the use of a “thin” access point, although such remains an imprecise term to this day. Thin, however, generally implies that significant functionality has been moved out of the AP and into the switch, in theory reducing the financial cost of the AP as well as the management overhead inherent in separate robust access points. This model further evolved into today’s controller-based WLANs (see Figure 1B), which allow access points to be connected to traditional wired-Ethernet switches, eventually coupling (perhaps even over a WAN connection) to the controller.
In order to fully illustrate and understand the structure of a given architecture, Farpoint Group uses a model based on the concept of planes, as follows:
- Management Plane – this defines how network management is performed in any given architecture. This function is almost always centralized and increasingly implement via an appliance. Such is the best way to achieve scalability as networks grow and assures configuration integrity, and is today almost universally applied by WLAN system vendors.
- Control Plane –defined here is how wireless network elements behave in real time, responding to data flows, error conditions, and other operational parameters. This function can be centralized in a controller, or distributed in each AP.
|
