Let’s be brutally honest. You can never really add performance or wireless reliability to a Wi-Fi network by using a controller.
In the wireless LAN space, vendors desperately try to differentiate themselves based on some unique “architecture.” But at the end of the day, does it really matter?
Different architectures use different technologies and different methods of getting clients connected to the network. It’s important to remember the obvious, wireless LANs are all about getting the clients connected to the network. What busy IT managers really want is wireless that just works!
The problem with fancy, complex and technical architectures is that they take the focus off what is really important – client performance and connectivity. Below is the good, the bad and the ugly of some today’s architectural choices:
Centralized WLAN architectures are designed to solve problems of control and security. Aruba was really the first company to perfect this architecture because they have a network-processing box. That’s what their controller is, a network processor. And it’s a good one. But it comes with limitations, high costs, encryption bottlenecks, and a ton of complexity. They recently announced support for distributed forwarding with a subset of their products. But they charge extra for this feature (and software licenses for almost everything else), and there are some severe limitations (eg. only 32 APs per controller in distributed forwarding construct).
Oh the ‘magic’ of single channel architecture (SCA). Invented and popularized by Meru, this type of system solves one problem: seamless roaming. The idea is useful for any application where roaming delays are a fundamental issue. SCA virtualizes the network to effectively look like a single AP to the client with all the APs sharing a single BSSID. The APs are simply “virtual ports,” and all the air protocols are arbitrated by the controller. This architecture is very fun to talk about and debate because it’s so unique. And you can do some amazing things by breaking the protocol. But ultimately, SCA quickly falls flat because it just doesn’t scale except in large expensive steps. But more important, it just can’t effectively deal with real-time interference or packet loss problems because it can’t determine the best signal path to send packets over.
Taking the controller out of the WLAN saves the costs of the controllers. OK. But it puts the ALL the burden wireless traffic and processing that traffic onto each of the individual APs. These APs now all must work properly in conjunction with each other for the network to work properly. Even without a hardware controller, the APs will need to be managed. This takes a SOME thing – either in hardware or renting a piece of software in the ‘cloud’. Though complex in the back end, these APs still have standard simple antennas and use only elementary omni patterns that are susceptible to all sorts of RF interference, both from their own APs as well as any systemic environmental RF.
Cloud-based WLAN architectures are designed solve the problem of controller complexity. This idea puts a virtual controller in the “cloud” and lets IT manager use controller services over the Web. Yet another innovative idea but what problem is it solving? And how does it really help with wireless part of wireless? Well it’s simpler to deploy. That’s good. But that’s about it. The APs are simple (ODM reference design), inexpensive SOHO-grade equipment with omni-directional, di-pole antennas that are subject to all the same problems that everybody else’s APs are subject to. This architecture has no performance advantage, is missing a bunch of enterprise-class features, doesn’t support multimedia, and is lacking opportunistic key caching. Again, also incapable of dealing with difficult enterprise RF issues.
Distributed forwarding architecture solves the problem of potential bandwidth bottleneck by not forcing ALL data back through the controller. This is a more flexible approach that allows for design adaptability based on client needs. This is probably the most practical approach, which is why many vendors are heading this directly. But for this architecture to be REALLY useful, it begs for smarter APs that can adapt RF signaling to clients as the environment changes. This architecture allows customers to put controllers wherever they want (at the central HQ, at each site, or even in the cloud) – or to not have ANY controllers at all by using a piece of management software that provides configuration, remote management and trend analysis.
So if it’s not about wireless architectures, what’s it all about?
It’s just about getting better Wi-Fi performance (that’s predictable wherever you are). There’s one distinct difference between wired and wireless performance; the physical way in which data is transported. The wire uses, well, the wire. Data, made up of electrical or light impulses travels down the copper or fiber and makes it to its destination with remarkable throughput and reliability.
But Wi-Fi uses radio frequencies (RF), electrical energy over the air, to deliver data. These radio frequencies must be shared (like a hub) by all users and are subject to interference and obstructions.
Because physics dictate Wi-Fi performance, there is a maximum amount of data that can be transmitted in a given amount of RF bandwidth (read up on Shannon’s Law if you’re interested in more specifics). Therefore, the key to true Wi-Fi performance in the air is always doing whatever is possible to maintain the maximum data rate without inducing unneeded signal (noise) into the air. Maximum channel capacity is then achieved by having highest possible signal-to-noise ratio (SNR) without having too much co-channel contention interference.
The single best way to deliver better Wi-Fi is to physically increase signal strength while at the same time decreasing the effect of noise. You’ve all heard about signal-to-noise ratio. And while SNR is very important, you must also factor in the interference into the equation. You might have great SNR, but if you have a high level of modulated interference (SINR), your clients will be spending much of their time queuing up, and waiting to contend for the medium with everyone else on the same channel.
Ultimately, with wireless, what matters most is the net throughput on the specific frequency (not some theoretical number on a datasheet).
So go ahead and look at the different wireless LAN architectures – they are a fun intellectual exercise. But when you want your wireless clients to just work, look for a system that does something to the actual wireless signals themselves to enable higher speeds and more reliable client connections.