I think that my younger self would be happy that some of the promises of AI and Machine Learning are being realised, after some years of being disappointed by marketing departments’ abuse of those terms. We are finally starting to see systems that start to automatically monitor, detect, and adjust themselves to correct problems. Ones that use data intelligently to automatically improve services, and ensure problem solving is not limited to “when the person can come around to fix it”.
It is with this context that I attended the recent Juniper Analyst Day to learn, in part, about the new capabilities of its Mist Platform and the Marvis AI solution and the implications of this technology for networks. So, let us breakdown the interesting aspects of this technology that Juniper were keen to showcase to the attendees.
Marvis AI is a network integrated machine learning platform that attempts to use the data generated by the network to improve services. Part of this is tied to Juniper’s Mist WAN Assurance platform that offers the sort of monitoring we have begun to expect from smart WAN and SD-WAN systems. The added layer is the built-in capabilities for the network to correct errors based on previously seen fixes and corrections and increased correlation between data and errors allowing for easier technology support. More info on this here: https://www.juniper.net/us/en/products-services/cloud-services/wan-assurance/
One thing I was interested to learn about was the conversational interface for this. Juniper is keen to give the example that you could just ask the platform “What network problems occurred last night?”, or “Who needs network support today?” and get an output that might function as a to-do-list for a network engineer or support staff. You can read more about this here: https://www.juniper.net/us/en/products-services/cloud-services/virtual-network-assistant/
I think the part of this that interests me most is the practical side of things, and reduced employee frustration and network issues and to discuss this I want to share a chart from the analyst day.
According to Juniper this chart illustrates the number of support tickets generated within the last year. Juniper has set up Marvis as the first port of call for network issues, learning commonly applied fixes and applying them before the user notices or they require support. Or when a problem is reported, attempting to let Marvis solve it before creating a customer support ticket.
Now the metrics above are interesting. On the one hand we might expect, with the growing transition to smart and more flexible WAN systems that can be configured and monitored remotely like SD-WAN, that the number of network problems encountered would drop anyway. However, this fall off in the context provided of 200% growth in organisations, and 280% growth in devices would imply that Marvis is effectively solving many problems before they become an actual technical support issue. If this trend continues it will be interesting to monitor, though it does raise questions about the levels of technical support staff needed at companies in the future.
One interesting current use case discussed on the day is tying Mist Intelligent Wi-Fi to individual identity tags in an office. This allows the location and movements of every employee to be tracked, so that if a COVID-19 diagnosis is found you can identify exactly where they were, who they were within 2 meters of and more. Of course, people might point out the ways that this can be used both positively and negatively by employers, but for COVID-19 at least it seems to be a useful tool to reopen offices.
It is worth noting that most large networking companies I speak to are pursuing this form of machine learning in some form. Velocloud Acquired Nyansa to help its journey towards self-healing networks and to harness machine learning and data earlier this year. Cisco have DNA but I have not seen any documentation linking that to Meraki or Viptela.
I want to leave a question in your minds about what this increase of automation will mean for service providers. Reduced technology support contracts? Less contracted support hours? Increased awareness of a customer’s business leading to service upsell? There are many knock-on effects of intelligent networks both positive and negative, and many more we have not even considered yet.
Where next for Mist? I would be surprised if they did not start to incorporate the location tracking and network awareness tools to create more security functions within the platform. In Cavell’s recent Enterprise Survey, we found that security was top of mind for networks eclipsing even cost, and interest in SASE models continues to grow. It would make sense for Juniper to build towards that direction, its competition already is.