We all expect a great Wi-Fi experience, especially when paying for a star hotel experience. However, most hotels are challenging to most Wi-Fi systems with guests and staff frequently roaming across halls, lobbies and courtyards with multiple Wi-Fi devices per person. A great Wi-Fi experience can help improve a hotel’s star rating or tank it. Don’t take my word for it, check out Yelp and Hotel Wi-Fi ratings sites (www.hotelwifitest.com). Bad Wi-Fi can be a curse.

Why? With staff and guests roaming around, devices frequently lose associations with AP’s and must reconnect to the Wi-Fi network. This can be a frustrating experience. If unresolved it leads to numerous support tickets for the hotel IT support staff.  In this blog, we’ll examine how traditional solutions recommend solving or working around this issue. We’ll then explain how KodaCloud takes an entirely different approach based on Machine Learning and AI techniques.

Let’s start with a smartphone. Virtually every guest has one of these. In some cases when roaming with a smartphone, it may take longer for the device (depending on the type, model, etc.) to switch to a better AP. This is because they use their own roaming algorithm. In other instances, users may have to hard reset the Wi-Fi connection. In the worst case, they would need to manually switch SSIDs.  Sounds simple, right? Wrong. Think over this. How is a guest supposed to know which AP offers the best signal when roaming at any given time? Or if that AP can assure the same throughput as the prior AP from whose range the guest is moving out of.  What’s worse, if the device is “sticky” to a specific AP, it is very unlikely that the device would gratuitously disconnect and reconnect to neighboring AP’s. Unfortunately, it must be forced to disconnect. (See a related article on this topic http://wifinigel.blogspot.com/2015/03/what-are-sticky-clients.html).

How is a guest ever expected to know of this setting? Such configuration nuances cause tremendous user frustration and support tickets with the hotel IT team. Who then investigate the issue independently (without context) and try to arrive at a reasonable fix. Often at the expense of a few hours or days’ worth of end user productivity. Toggling the Wi-Fi SSID on the smartphone is also a poor end user experience. That’s because applications that were active when connected to the Wi-Fi network would have to reestablished. Surprisingly, there’s never been a convincing solution to this. Off the shelf AP’s simply lack any configuration knobs to manage roaming.  This means when deployed in a hotel with a several hundred thousand square feet of space, users will almost always experience Wi-Fi connectivity loss. Some enterprise Wi-Fi solutions offer proprietary “steering” of clients to APs. But their systems are hardly intelligent. These solutions have too many  parameters to consider, rely on fragmented data and lack historical context. They are not nearly as efficient an AI system that leverages a large data set to make decisions and learns from past behaviors.

KodaCloud takes a very unique approach to solving this issue. In addition to utilizing the standard  802.11 k/v/r methods as needed, our Artificial Intelligence (AI) system determines which among the known AP’s should be designated as “preferred AP” at any instant. This decision rests on historical or current data and is based on usage and signal quality among others. Having richer context through Machine Learning and data mining helps make (and validate) the most optimal AP decision. Furthermore, using machine learned fingerprinting, our system identifies clients that are sticky. The Machine Learning (ML) system uses different steering parameters and mechanisms to steer clients to the preferred AP.  Finally, to minimize excessive client roams, our Machine Learning (ML) system also identifies stationary clients in the pool and chooses not to steer them, unless absolutely necessary.

The ultimate goal is to achieve a healthy mix of steerable and stationary clients by intelligent load balancing across APs. The analysis extends beyond client roaming. If a network is largely misconfigured, such that the furthest AP’s in a coverage area are broadcasting at the highest power and a client device associates with one of those APs, the closer APs are blindsided. This leads to spatial reuse problems and reduces the effective capacity for everyone on the network. We overcome this situation with our patented Machine Learning algorithms, that ensures client roaming is as non-disruptive as possible.

roaming

Here is an example from our own network of how roaming scenarios are successfully optimized by AI to achieve the best outcome for every client in the network. Tell us how you’re solving Wi-Fi roaming issues for your guests and the hotel IT team.

Nilesh Savkoor

 

Chat with a Wi-Fi ExpertAsk us a question