Does Wi-Fi Architecture Matter In 2018?

Last week, I looked at how to avoid RF interference in WLAN deployments. While this is something that affects everyone that uses Wi-Fi, the architecture your network uses is something that you can control (based on which vendor you choose).

One of the primary differences between WLAN architectures is where and how the system is controlled. Broadly speaking, some vendors are controller-based and some focus on distributed intelligence and are considered to be controller-less. The reality can be more complex, as you’ll see below.

When Wi-Fi was first introduced in the late 1990s, access points were standalone, isolated pools of connectivity, mostly designed to provide access to the Internet. In the early part of 2000, it became obvious that Wi-Fi must be somehow joined to the wired network. At that point, however, the processing power required for access points to handle networking decisions independently was overly expensive. The result was a centralized controller, in which access points were distributed throughout a facility and backhauled to one or more controllers. The controller was then connected to the wired network.

Complex decisions like firewalling, access control, authentication, roaming, and more were handled by the controller. This required traffic to travel from the AP to the controller and back, but at the time this lag was not a significant barrier to adoption. Management and control were centralized, and Wi-Fi joined the network.

Centralized controllers were not without their drawbacks, however. Because of the nature of their connection, each controller model was “capped” at a certain number of access points. This did not seem like a limitation until it was time to purchase the AP that was “one too many,” at which point an expensive new controller also had to be purchased. The controllers themselves became a single point of failure, requiring the purchase of redundant controllers in areas where Wi-Fi was considered mission critical. This centralized controller architecture is still in existence, although the need for it has disappeared.

Over the last decade, the cost of processing power has fallen sharply, largely due to the advent of affordable, powerful mobile clients. Some WLAN vendors that originated at this point in the continuum built their products from the ground up on highly capable processing power that had become economically available to all. These vendors were able to build a networking system that did not require a centralized controller to function. In this model, access points were able to communicate with each other in a fashion similar to the way routers communicate on the Internet: control became distributed.

Management remained centralized, resulting in a much more cost-effective model that was easy to scale and equally easy to manage.

These two architectures—centralized controller-based and distributed controller-less—form the extremes in control for WLANs. As the distributed controller-less architectures gained popularity, several leading controller-based vendors created or purchased their own controller-less WLAN offering, while continuing to sell their controller-based products. The controller-based model has come under increasing fire in recent years due to several issues, including:

  • A client may go from one access point’s coverage area to that of another, sometimes crossing Layer 3 boundaries and into another subnet as it does so. As the client crosses into another subnet, its signal will be received by an AP that does not share the client’s IP addressing sequence. Any delay or jitter, which can happen when traffic has to transitthe network to the controller and back, can become problematic, especially for sensitive transmissions like voice or video.
  • As APs become capable of much greater speed, the amount of traffic that would have to be backhauled to the controller becomes increasingly untenable. If you architect a network to forward data to a central control point, as it is in the controller-based model, there is no way to balance multiple gigabits-per-second of data across multiple controllers. In addition, some vendors require a controller upgrade to support 802.11ac. As we look ahead to 802.11ax, will your vendor be able to mix and match older APs with 802.11ax APs without requiring costly hardware upgrades?

All posts in this series:



Alexandra Gates is a Senior Product Marketing Manager at Aerohive Networks, where she helps define market strategy and vision for the cloud and WLAN products. She is a CWNA with a comprehensive background in wireless technology, including capacity and management planning, RF design, network implementation, and general industry knowledge.

Leave a Reply

Your email address will not be published. Required fields are marked *