Intelligent Wi-Fi
Written by: Sam Fenwick | Published:
Thorsten Freitag, VP EMEA of Mist

Thorsten Freitag, VP EMEA of Mist, talks to Sam Fenwick about how his company’s use of machine learning is making it easier to manage and troubleshoot Wi-Fi networks

Anyone’s experience of using Wi-Fi knows that service at large venues can be patchy and a host of different things could be the problem, making it difficult to troubleshoot, while the sheer number of possible interactions between different devices and access points means that trying to replicate the problem can be a fool’s errand.

Mist, a company hailing from Silicon Valley, has its sights on changing this through using cloud-based machine learning. Thorsten Freitag, its VP EMEA, tells me it was founded when a Fortune 100 company had just deployed a wireless network, only for its CIO to say: “This is great, but on Monday, how do I know what experience I’m giving my users?” The company was created to solve this issue using artificial intelligence back in 2015, and it has been providing what it claims to be the industry’s first learning WLAN for about 18 months.

In that time, it has picked up about 250 customers, with one in 10 hailing from the Fortune 500, and these include “the top e-commerce company, the top internet search engine company, and the top social media company”.

Now that it has established itself in the US, Mist is now looking to enter the European market, and that’s where Freitag comes in. He explains that it currently offers four services: two for Wi-Fi and two for Bluetooth.

Starting with Wi-Fi, Freitag says Mist’s Wi-Fi Assurance Service allows the person in charge of a Wi-Fi network to set service levels for user experience, and also offers dynamic packet capture. He demonstrates the system to me using a set-up that is monitoring the Wi-Fi network for Mist’s corporate office over in Cupertino, California.

The service levels that can be set include a range of parameters such as throughput, time to connect, roaming and coverage, and Mist’s system can tell the operator what percentage of the time those targets are being met and can analyse trends, giving it the ability to say whether certain operating systems or devices are experiencing the most problems.

The dynamic packet capture capability allows the system to do a “network rewind” back to the moment that a user experienced a problem and see what caused it, eliminating the need for a support person to go on-site with a packet sniffer to try and recreate the issue. Freitag says these capabilities mean Mist’s customers have been able to cut the costs of operating their Wi-Fi networks by 40 per cent.

The second and most recent service is the Virtual Network Assistant (VNA). “It adds natural language processing [capabilities], so you ask simple queries, such as ‘how did wireless client Randy’s iPhone perform last week?’. It aggregates the information so a helpdesk person who isn’t an RF expert can understand what’s going on and make a conclusion.”

Returning to the earlier example, it might report that ‘Randy’s iPhone’ had three problems, and it could also let the user know whether any of these are site-wide issues. Freitag adds that a user can also click an ‘investigate’ button to see the data that the VNA used to make its conclusions and provide them with the information they need to troubleshoot the issue. It also allows users to run queries for entire sites or quickly find out the number of particular devices (such as iPhones) on the network.

Curing the BLE blues
Turning to Bluetooth, Freitag says Mist’s Wi-Fi access points include a 16 Bluetooth Low Energy (BLE) antenna array, making each equivalent to eight normal BLE beacons (it also sells an access point with just BLE functionality), and Mist’s access points have a port to connect them to IoT devices. The BLE-enabled access points are combined with unsupervised machine learning algorithms to provide “one- to three-metre location accuracy 90 per cent of the time” and BLE coverage to entire rooms. Freitag shows me how the Mist platform handles BLE: a floorplan overlaid with blue head symbols – “people with phones using our SDK” – and green icons corresponding to third-party devices.

With the standard approach to BLE, if someone wanted to adjust where someone received a notification via BLE, they would have to physically move the BLE beacon. Freitag says Mist’s system allows users to create virtual beacons instead.

He says Mist’s services include BLE Engagement, which covers pushing messages to people to boost engagement, give them directions or proximity notifications; and Asset Visibility, to make it easier to locate physical assets such as wheelchairs in a hospital or pallets in a warehouse, or even security personnel. And Mist can use its BLE system to provide venues with location-based analytics data.

“So if you’re a hotel or a convention centre and you want to know which areas are getting the most foot traffic, how many people are visiting that area, and what the maximum and average dwell time is, that’s all built into our system as well,” Freitag adds.

Freitag says the prime use-cases for BLE-based location services are in retail, healthcare and hospitality. “Retailers, for example, want to greet customers when they come on-premise and engage with them using personalised ads. But this has been hard using battery-powered beacons – they’re expensive to deploy and difficult to manage. For these customer loyalty programmes to work, Mist virtualised the BLE experience, eliminating reliance on battery beacons.”

Trying to kick the AP habit
Freitag says Mist would prefer not to be in the wireless hardware business, but it’s necessary to enable it to gather the data that underpins its services. For example: “We rewrote the control plane for 802.11 so we collect more than 100 states from every mobile client every two seconds – that’s metadata, not payload information – so we can look at trends and do troubleshooting.”

Mist’s goal is getting to the point where Wi-Fi networks using access points from other vendors can access about 80 per cent of the same capabilities (including being able to set service levels and perform dynamic packet capture) available when Mist’s cloud platform is used in conjunction with its own access points.

Freitag adds that in addition to special access points, a purpose-built cloud with AI is needed. Mist developed this from scratch with the same technologies used by Netflix, LinkedIn and Google “for the scale, resiliency and agility that come from a micro services architecture”.

While at first glance the natural home for this service would be public venues, where the Wi-Fi network operator can’t rely on enterprise mobility management software to gain insight into the majority of the devices being used, Freitag says it is being put to use in some carpeted enterprises to allow them to see trends from the wireless network’s perspective.

Hello, Europe
Mist’s European operations have been up and running since October 2017 and grown to about 10 people, with the UK having a dedicated salesperson. However, as is common with many companies in this space, it has its own sales channels in the form of “distribution partners, resellers and trained engineers”. Freitag adds that Mist resellers need experience with enterprise Wi-Fi and “access into large enterprises, finance, education, government, retail and should have experience with those vendors that have been in the market for a few years such as Aerohive and [Cisco Meraki]”. Freitag adds that those resellers with a managed service provider mentality are attractive to Mist as the company is both service-focused and cloud-based, which gives such companies the opportunity to transition towards offering a wireless service.

Because the main skill required by a reseller is surveying and knowing how to deploy Wi-Fi access points, Freitag says: “Our specific Mist training would not take probably two to three days for the engineers – it’s mainly about how to use the cloud-based system and how to access the data, if there’s any customisation required, if there are any applications that want to access the data that we capture.” He adds that Mist can assist its resellers with packaging services around wayfinding and how to use its open APIs to deliver services.

In a world in which many wireless start-ups could be accused of being a solution looking for a problem, it’s refreshing to find one that was founded to address a clear need. I can’t help but wonder where else machine learning could be used to make life easier for wireless comms users.

CV – Thorsten Freitag
In addition to his role as Mist’s VP EMEA, Freitag is also the CEO and founder of Valesco Ventures, which started in July 2016 and builds and grows international sales subsidiaries for technology companies in the IT infrastructure and security markets. Before then, he spent more than two years at Infoblox, a critical network services provider. He joined Infoblox after spending five years as senior VP EMEA at F5 networks.

Prior to that, he was a senior VP at Siemens Enterprise Communications and also spent nearly three years as the founder and managing director of the Business Acceleration Group, which supports companies in the information and telecommunications markets.

Freitag joined Cisco Systems in 1995 and left in 2004, having served in a number of positions including head of sales and COO at Cisco Systems South Korea and managing director of its operations in Sub-Saharan Africa.


This material is protected by MA Business & Leisure Limited copyright.
See Terms and Conditions.


Comments
Name
 
Email
 
Comments
 

Please view our Terms and Conditions before leaving a comment.

Twitter

Land Mobile is the only monthly publication exclusively dedicated to wireless communications for businesses. Launched in 1993, this leading industry title provides practical advice, expert opinion and commentary and insightful, informative, truly authoritative editorial.

St. Jude's Church,
Dulwich Road,
London,
SE24 0PB,
United Kingdom

MA Business is part of the Mark Allen Group.

PMRExpo 2018 show guide

I would like to

Or Click here
to visit the website