Follow me:

Mist @ Mobility Field Day 4

Mist is now regularly presenting at Mobility Field Day. They often like to talk about some of their success stories and update us on their “travel to the self-driving network” (i.e. how their Artificial Intelligent system is evolving). This year was no exception!

If you want to catch back with the presentations and watch them, here are the links you need:

In this episode, we give our point of views on the different topics Mist presented on.

For more insights form the other delegates, please read the following articles:

Wi-Fi 6

I was personally very pleased that they had a Wi-Fi 6 AP running when we showed up in the morning for breakfast. They even had a live view of the spectrum analyzer on the projector show casing OFDMA. I thought that was pretty kool! After adding a couple more Wi-Fi 6 devices to their network, we worked with the guys from Mist (Wes and Robert) to generate even more traffic on the channel. Here is a look at it:

Later on, Wes Purvis presented a few slides on their new Wi-Fi 6 AP-43. We also got to hold and see it in person. Hopefully, we will get to test it out soon. One of the key takeaway for me is that Wes showed us how to validate that the AP is using OFDMA. You have to look out for a frames with a frame control of 0x2400 ;).

Update of the Mist AI

We talked a little bit about it with Bob Friday when Rowell interviewed him for the podcast a couple of episode ago (https://www.cleartosend.net/20-years-of-wifi-bob-friday/). During Mobility Field Day, Bob Friday and Sudheer Matta gave us even more details on how the Mist AI is evolving.

Mist developed, what I would called a virtual assistant named Marvis that help their customer in managing, monitoring and troubleshooting their infrastructure. The principle is easy, you just ask Marvis a question (very similar to Alexa, except that you have to type it for now), and Marvis uses artificial intelligence and machine learning to give you an answer.

This is not new. What is new though, is how better it has become at detecting anomalies. Mist is now saying that Marvis is able to find anomalies 99% of the time. This is not easy to achieve! They were able to do it using special algorithms based on Recurrent Neural Networks.

The way the anomaly detection work is as follow:

  • The system looks back at what happened on the network for the past 2 months taking into accounts time of the day and day of the week
  • Using that data, it predicts what the next hour will be
  • If there is a deviation, then it will detect an anomaly

A couple of new features:

  • Switch health view
  • AP health
  • Bug detection and auto-patching fix
  • Marvis giving things to do when you arrive in the morning at the office

Mist Edge – Microservices to the Edge

They also talked about one of their new products called Mist Edge. With Mist Edge, they are bringing their Microservices architecture in their customer networks. For now, It solves the problem of having to centralize VLANs somewhere in the network. This allowed Mist to win some larger campus deals as they are traditionally based on an architecture where all the traffic is tunnelled back to a centralized controller in the LAN. But in the future, they could bring other microservices to the edge.

Managed in the cloud. The AP discover the edge automatically when you connect it to the network. You can tunnelled per VLAN and L2tpv3 is the protocol used to tunnel the traffic between the APs and Mist Edge.

It was a must for Mist to address the larger campus part of the market. Now they have a solution that will help them sell in large campus environment (higher education for instance).

Location

Bob Friday was very keen on giving us an update on their location services. As you can see in the picture of the inside of the Mist AP, there are a few BLE antennas. This allows Mist to perform indoor location.

Since Google and Apple are embracing the BLE technology for indoor services, Mist believes that it will be important to have an infrastructure that can leverage it. What is interesting with Mist is that they do it without having to install an overlay network and they used AI and ML to learn the path loss model for every phone and every AP.

When they presented about it, they had a slide that say the “Machine Learning eliminates surveys”. People on Twitter took it out of its context and thought that they were talking about Wi-Fi surveys. However, they were talking about location services using BLE.

Juniper

I think it was a little too early for Mist to present they are planning on doing with Juniper. Hopefully they will be back in 2020 to MFD 5 to talk about it.

There is a lot of potential. We could see Mist starting to add Juniper switches and routers to their dashboard and integrate the data they gather from these equipment to sharpen their algorithms and include more LAN analytics.

Wireless Network Engineer and Owner at SemFio Networks. CWNE #180. Living in London ON Canada, born and raised in Dijon, France.

Join the discussion

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

More from this show

Episode 183