In Wi-Fi They (Don't Really) Trust

By scottm32768 posted Jan 11, 2016 08:01 AM


Sometimes, the biggest problem with the network is its very existence. Anytime something breaks, the fingers start pointing at the network. Database stopped responding? It must be the network. Client can't access the Internet? Must be the network. Never mind that what the client can't access is just their home page and everything else is working…


The problem isn't so much that the network exists, but that it exists and most users, and even most IT pros, don't understand it. Now we take that complex system that people already have a difficult time understanding and replace the simple Cat5 cable with… Magic? Arthur C. Clarke once wrote that any sufficiently advanced technology is indistinguishable from magic. For many people, wireless is a magical black box. Actually, it's usually an opaque white box, but that's beside the point. Things happen in it, but they can't be seen and they are not easily understood. The explanations for how it works, or more likely why it doesn't work, generally involve lots of vague hand waving motions and end with either blaming the client or the network, depending on which side you are on.


Now when something breaks and there's nothing obviously wrong with the device people trust, it's logical (from their perspective) to blame the thing they don't understand. It's known that it needs to be working for them to do what they want, so that must be what's broken.


This phase in a problem is where we start measuring the Mean Time To Innocence (MTTI). This happens regularly enough that we have a parody acronym for the time it takes to show it's not the network. Unfortunately, this is more difficult to achieve in Wi-Fi than with wires because using RF instead of copper for layer 1 brings in a whole slew of new variables and failure modes. To make it more difficult, Wi-Fi problems involve highly variable clients that may belong to random guests and be locked down so they are difficult to troubleshoot. Oh joy!


My original point was going to be that the network is not trusted because the network is not understood, but then I started thinking about it. People don't really understand how their computer works. Most don't understand their car, for that matter! Yet these things at least can be trusted. The key is that trust is earned.

It would appear that most Wi-Fi installations haven't earned that trust. It might be due to client issues, but the users don't know the difference, nor should they really be expected to. Unfortunately, most of the problems we face in Wi-Fi stem from the crazy variety of client implementations.


In order for the wireless infrastructure to be trusted, the wired infrastructure needs to be a solid foundation to build on. If the wired infrastructure isn't stable, you really need to fix that first! If that's an issue, hopefully you are either on or have a good relationship with team with that responsibility. Part of it might be a need to build in high availability features, but in most cases it seems the bulk of problems in the foundation are due to a lack of standardization and poor overall design. The same is true for the wireless network.


The problems we have with clients usually aren't the real problem. They are a symptom of the real problem. A lack of understanding of proper wireless design. Everything comes out of the design and the design is all about RF. Layer 1. The physical layer that has replaced the copper wires. The part that in a wired network either works or doesn't, and if it doesn't you swap out the cable. This is not an option we have.


We need to be careful with shortcuts, too. The predictive "survey" might give you a starting point, but if someone doesn't validate that after installation, you can't know the design is actually working correctly. If possible, start with a proper site survey first. You have a much better probability of having a good design to start with. This is a lesson I've had to learn myself. As I've become better at gauging how a predictive survey will map to the real world, I've also come to better understand the limitations. Without a survey, you have no idea what the RF environment is like. For example, if you find the noise floor is much higher than average, you may not have enough APs or may have to run them with more power than expected. Better to know beforehand and get it right the first time, than have to go back to fix it.


The way for the network to earn the trust of the users is for it to be designed correctly for their needs. This means the user will have reliable connectivity they never have to think about, so their first reaction when something doesn't work won't be, "Is there's something wrong with the Wi-Fi?" It will be, "Is there something wrong with my device?"