Waystream initiates research project in Machine Learning / Artificial Intelligence

The Swedish technology company Waystream that provides equipment for fiber networks has initiated a research project on machine learning and artificial intelligence in relation to operation of broadband and city networks. The aim of the project is to evaluate various methods for machine learning and artificial intelligence to be used in the operation of fiber-based broadband networks in order to improve availability and reliability for end-users.

The research project has been initiated by Waystream and consists of participants from several organizations. The RISE research institute is part of the project and contributes with expertise in artificial intelligence and machine learning. Swedish network operator Lunet, which operates a city network in LuleƄ, also participates in the project that is led by Johan Sandell, CTO at Waystream

"Artificial intelligence and machine learning create new opportunities for analyzing large amounts of data to find deviations from what is normal. Applying this to the type of information provided by Waystream's switches through telemetry allows us to find errors that affect network traffic much faster than before. For a city network operator it means that you can get alarms about anomalies in the network that you would not otherwise have detected before it already affects large end-user groups. So the goal is to investigate new tools for network operators to use to locate and even predict errors and take action that reduces the impact on end-users.", says Johan Sandell.

Waystream introduced telemetry technology in its products at the end of 2019. By using telemetry, operators can collect thousands of operational parameters from every switch, every minute. These parameters provide details about the network performance. In a large network there can be hundreds, even thousands of switches. Active service assurance and monitoring thus create a unique insight into how the network actually works.

"It's a huge amount of data available for analysis. More than what a human alone can keep track of. At the same time, each parameter conveys important information. Therefore, machine learning, where the systems learn what is normal at any given time each day can be a powerful tool. When any of these thousands of parameters deviate from the normal, you can do an automated analysis and alert the operating staff if necessary. That is the goal", explains Johan Sandell.

The city network operator Lunet is part of the project. Among other things, they provide collected data that shows both normal conditions and actual events that affect end-users in their network. Daniel Henriksson at Lunet works daily with the operation of the network.

"We see a great potential in this technology. We work daily with different types of problems that affect our services and the end-users' experience. Sometimes there are problems on the Internet that affect our users. Then it is useful for us to quickly determine that the root cause is outside our network so we do not spend unnecessary time on troubleshooting. Likewise if an issue is very local, maybe affecting only a specific neighborhood in the network, then we want our customer service and affected users to have correct information quickly. That's why this project is so relevant to us", says Daniel Henriksson.

The research project is supported by Vinnova as part of an initiaitive to expand use of machine learning technology to Swedish companies. The technical expertise in this area comes from the research institute RISE. Dr. Rami Mochaourab is an expert in machine learning. He says;

"We are a team within RISE participating in the project. For us, this is an exciting application, to see how different machine learning techniques can be used on telecom networks and, not least, city networks, which are an important part of the modern communication infrastructure. There is a bidirectional knowledge transfer here that also benefit RISE as it increases our understanding of fiber access and city networks. ", says Rami Mochaourab.

The research project extends into 2021, but already after the summer 2020, the first results will be presented to the participating project organizations.