How AI could help Wi-Fi detect happy users – and track them

wi-fi AI
Image credit: PlargueDoctor [sic]

Wi-Fi architectures need to leverage AI and the cloud to take the wireless LAN to the next level, add value and – if nothing else – make life easier for IT managers.

So says Juniper Networks, who officially entered the enterprise Wi-Fi market back in June – two months after completing its $405 million acquisition of Wi-Fi start-up Mist – with two new Mist products: a Wi-Fi 6 access point and an edge device called Mist Edge. The latter box extend Mist’s cloud microservices app environment to on-premises environments, which enables IT managers to centrally manage everything for local and remote campuses.

However, the Mist platform comes with a few key extras: an AI engine, support for Juniper’s microservices and location technology that could enable highly accurate location-based apps on Wi-Fi networks.

Juniper is well aware the enterprise WLAN market segment is already pretty crowded, but Juniper senior system engineer Charles Cheang argues it’s a segment that hasn’t seen much innovation from an architectural point of view for at least ten years. The problem, he says, is that the WLAN architectures that dominate the market today were developed before the rise of smartphones, public cloud services like AWS, GPUs and network virtualization.

“What’s missing is a modern Wi-Fi architecture that can effectively take advantage of the virtualized cloud environment that enterprises are moving towards,” Cheang says.

Making Wi-Fi user-centric

Cheang illustrates the problem by pointing out that Wi-Fi management today is network-centric rather than user-centric. “If someone walks into a hotel, and they have a bad experience with the Wi-Fi and they complain, the typical IT department response is to pull up some dashboards and say, ‘Hey, the access point is up, so you should be having a good experience’. But this is not the whole story  – ‘up’ is not the same as ‘good’.”

Meanwhile, he adds, the onset of Wi-Fi 6 promises better throughput, performance and capacity for Wi Fi networks, but it also introduces new complexities. Mist’s AI engine will enable automation for its Wi-Fi 6 APs as well for things like intelligent load balancing between radios, service levels for OFDMA subcarrier assignments and basic service set (BSS) coloring assignments for high-density Wi-Fi environments.

It’s worth noting that Juniper isn’t the only vendor applying AI to Wi-Fi. For example, HPE subsidiary Aruba announced  new Wi-Fi 6 access points earlier this year that leverage AI and machine learning to automate RF optimization for better channel planning and power planning.

But what Juniper really wants to do with AI and analytics is to make Wi-Fi management less network-centric and more user-centric by measuring KPIs that reflect the user experience – in other words, not just knowing if the network is up, but whether the user is happy with the connection they’re getting.

“Traditionally, when you’re looking at the user’s connection, you just look at CPU utilization, bandwidth utilization or whatever metric is available from an infrastructure perspective,” Cheang says.

By contrast, he says, Mist’s predictive analytics and correlation engine (PACE) measures more than 180 metrics every two seconds, crunches the data in the cloud and produces the results across seven “services level experience” (SLE) metrics that reflect the user experience, including time to connect, throughput, coverage, capacity, roaming between APs, successful connects and AP uptime.

IT managers can use Mist’s “Marvis” virtual network assistant to sort through all this and ask questions like “show me unhappy users” and get an anwer based on the seven SLEs. As Marvis keeps learning, it can eventually predict which users are about to become unhappy, Cheang says.

Better indoor tracking

Meanwhile, another key capability Mist claims to bring to the table is more accurate indoor location-based services.

Cheang explains that while mobile mapping apps that leverage GPS and cellular neworks are reasonably accurate outdoors, indoor accuracy is another story. “Indoors, the only two ways to do location services are Wi-Fi location, which doesn’t really work great, or Bluetooth, which does work but isn’t scalable because you need battery beacons every 15 to 20 feet.”

Mist tackles this with its proprietary virtualized Bluetooth Low Energy (vBLE) technology, which – as the name suggests – virtualizes Bluetooth beacons in Mist APs outfitted with 16-element BLE antenna array access points.

The result, Cheang says, is accuracy of between one and three meters, and lowers the latency so that the blue dot on the map shows your location in real time rather than lagging ten seconds or more behind your actual position.

This capability opens up all kinds of new service possibilities for enterprise Wi-Fi, Cheang says. “For example, when guests arrive at a hotel, the Wi-Fi network can greet them at the door, facilitate check-in, give them turn-by-turn directions to their room, and push coupons for the hotel restaurant or gift shop .”

Other examples include retail stores where the customer could use an app that summons a sales rep by pushing a button that would alert the sales rep’s device and show the customer’s exact location in the store.

Cheang demoed this app during a media briefing in Hong Kong last week, and while it worked, implementing it in real life would require convincing customers to download yet another app on their device and give it permission to track them all over the store.

Less is more

Apart from the whiz-bang AI-powered capabilities, Juniper also says that there are practical benefits like cost savings realized mainly from simplified operations and automated troubleshooting (which ultimately means less support tickets for the IT manager to deal with as the Marvis assistant learns which problems can be solved automatically and which ones need to be passed on to the human support team), as well as the increased productivity you get from a functioning WLAN.

If nothing else, says Margaret Lee, Juniper’s country manager for Hong Kong and Macau, reducing the complexity will take some pressure off IT managers trying to keep up with their enterprise’s escalating broadband demands.

“With increases in bandwidth, the ever-growing number of IoT devices and applications, IT managers are dealing with more and more complexity, and there’s no longer so much money to employ tons of IT guys, so they’re expected to do more with less,” Lee says. “So what they are really looking for is automation and artificial intelligence.”

Be the first to comment

What do you think?

This site uses Akismet to reduce spam. Learn how your comment data is processed.