During the Data World event, Erwin Halmans, Project Manager of Data-Driven Network Operations & Assurance at T-Mobile Netherlands, Shounit Lax-Swisa, CEO of Company Booster, and Anodot Chief Data Scientist and Co-Founder Ira Cohen got together to discuss the future of network monitoring.
Lax-Swisa, who has worked at the intersection of technology and telecom for many years, highlighted the importance and relevance of the topic: “We’re living through a very fascinating era. Some will call it the 4th Industrial Revolution, others call it a digital transformation. The bottom line is that we see many new cutting edge technologies pushing the market forward and creating interesting opportunities and use cases. AI and machine learning is one of those technologies.”
AI/ML have been gaining traction in the telecom industry in the past five years, specifically for use cases such as fraud prevention, security, revenue assurance, billing, marketing and many others. The main challenge for telcos in the past few years, however, has mainly been demonstrating exactly how this technology can contribute to their overall business value and bottom line.
How do you monitor your network today and what are your main challenges?
Halmans, who manages network monitoring at T-Mobile Netherlands, specifically for the fixed network of 4G mobile phones, said, “The biggest challenge for us in terms of monitoring is that we have a lot of legacy networks. The architecture and landscape is very scattered, which means that we have a large variety of monitoring tools throughout the organization to monitor the status and performance of the network.”
Additionally, there’s difficulty in acquiring a holistic view into the overall performance and stability of the network. This can be attributed to the array of devices, technologies, and vendors, as well as the increasing demand from enterprise customers for advanced services.
Can AI/ML be used for network monitoring?
As Ira Cohen explains, the short answer is yes. The challenges in network monitoring that T-Mobile and other telcos face are well suited for machine learning.
“The biggest reason is the notion of scale,” Cohen said. “With tens of thousands of devices each outputting sensor data, as well as monitoring other types of data such as services, the data center, and so on, the complexity is very high.”
With the speed at which decisions need to be made within a network, the telecom industry has always been at the edge of automation. Machine learning provides the granular visibility and the ability to scale. A huge leap from the brute force approach that teams have traditionally tackled monitoring challenges.
What are the main challenges with AI implementation?
According to Halmans, his team has confronted obstacles with integrating AI into the day-to-day operations of the business.
”One of the challenges we face is how to obtain or extract data from all the various devices and transform them into the same format so that we can analyze,” Halmans said. “Don’t underestimate the amount of work that goes into this.”
Additionally, telcos often try starting projects that are simply too large to generate value within a reasonable time frame. Erwin recommends starting as small as possible and then to increase in complexity from there. He solved this by identifying the single problem causing the largest amount of issues in the network. From there, he and his team created a project around that issue and worked towards a solution: “Once you succeed in the first step, enthusiasm in the organization also increases.”
What will the telco landscape look like in the next few years?
There is a clear trend towards the digitalization of networks with a software-oriented approach, which is inherently more flexible and dynamic. This flexibility does increase complexity to the network, which creates the need for a more proactive, automated monitoring solutions powered by machine learning.
T-Mobile is also already in the process of rolling out 5G, although the number of devices is currently limited. They will see this number explode in coming years as 5G and IoT applications proliferate. At that point, it will be impossible to monitor networks manually, resulting in further automation and increased visibility into each networks’ devices and services.
For more tips and insights, watch the panel on the future of network monitoring.