A few weeks ago I wrote a blog on the evolution of wireless LAN and the alternate architecture that Meru offered. Since then I've had updates from many wireless LAN solution providers and one of the more interesting vendors was a company called Aerohive.
A typical wireless LAN solution today utilizes "dumb" or "thin" access points and a centralized "controller" that handles all of the intelligence and configuration parameters of the wireless network. This architecture was put in place when wireless LANs started to grow in popularity and "fat" APs could not scale because of the complexity of managing a bunch of independent APs and the lack of system-wide intelligence. The controller re-introduced Wi-Fi as a cohesive system (rather than a collection of independent parts) and brought centralized management to wireless LAN, allowing it to become a mainstream networking technology in companies. However, as I pointed out in my previous wireless LAN blog, we are on the precipice of another transition as companies are looking to make wireless the primary networking technology instead of a convenience-based augmentation to the wired network.
Before I get into the specifics of Aerohive, let me walk through what the limitations are of a controller-based solution. The controller can be thought of as the "brains" of today’s de facto standard wireless network. All the important "control" functions such as channel selection, QoS and secure fast roaming are executed by the brains. Typically all data traffic going to and from the wireless devices must also pass through the controller for processing and forwarding. Within a building, prior to 802.11n, this wasn't all that big of a problem. However, with the emergence of 802.11n and when you want to extend the wireless network over the WAN, the controller can act as a bottleneck and become a single point of failure.
Since all traffic must typically pass through the controller, all wireless traffic in the branch or remote location must traverse the WAN, go back to the controller and then come back to the branch, effectively crossing the WAN twice. The obvious solution to this problem is to put a controller in the branch, an expensive one, especially if you need redundant controllers for high availability. More recently, in order to improve performance and reduce cost, controller vendors have developed hybrid approaches where the remote APs forward data locally but control functions or intelligence is provided by a central controller. Ultimately this is a tradeoff between performance and cost versus reliability, as the controller and also the WAN connection, are single points of failure for the functionality and intelligence of the remote APs.
Aerohive’s solution has no controller as part of the solution. Instead of having thin APs, the solution uses APs that have intelligence but can be managed as a single entity. The intelligence and management are distributed across the "hive" continuously. The actual management itself is done through a centralized console and then the configurations are pushed out to all of the APs. The APs then all work together to control features such as stateful roaming, load balancing, RF control, security features, etc. In a sense, the wireless "hive" operates a lot like current wired networks where the control is distributed across a network of routers. Routers can be centrally configured but once the configuration is pushed to all of the routers, the control of the network is distributed across the entire network. That's the only way large WANs can scale, be manageable and maintain a high level of performance.
The net result of the solution is a high performance wireless network at a much lower cost than if the deploying organization had to buy pairs of controllers for every location. Also, because control is distributed across all the devices, there is no single point of failure. Again, the analogy of the router-based network holds true, as the loss of any one router does not impact the network, just that one device, so the resiliency of Aerohive’s solution is very high.
As far as I can tell, Aerohive is the only vendor that has moved to a fully controller-less architecture. There are some vendors that have eliminated the need for controller hardware by moving the software into a VMware appliance or into the "cloud," but a controller is still needed in the solution and it still requires a constant connection to the AP to work as an intelligent enterprise class AP.
The evolution of end user devices coupled with 802.11n speeds has given enterprises an opportunity to move to a predominantly wireless network if they should so choose. This has created a transitional moment in the evolution of wireless LAN technology for smaller vendors to create innovative solutions to help organizations achieve this goal. I'm expecting 2011 to be a very busy year for wireless LAN as both small vendors and the market leaders (Aruba and Cisco) continue to evolve their products to help companies move to a network where the majority of access is wireless.
Wireless isn't just for ad hoc connectivity any more. Companies are running VoIP, video and other real time applications over wireless LAN now, and the choice of wireless vendors shouldn't be based on who the wired vendor is. Network managers should evaluate a wide variety of vendors with the most demanding applications to find the right solution.
