Big news: Tryolabs joins Qubika. Same teams, broader set of combined capabilities. Read now ->

Case study

Weather-related outages prediction

The time interval between the outage detection, its diagnosis and labeling, went from 20 minutes to less than 1 minute.

Industry: Telecommunications
Client
Solution: Predictive analytics

Effective outage prediction

SES delivers global content connectivity solutions and is one of the most important satellite operators in the world. SES was the first operator to deliver a differentiated and scalable GEO-MEO offering worldwide, with over 50 satellites in Geostationary Earth Orbit (GEO) and 20 in Medium Earth Orbit (MEO). We teamed up with them to build a MEO monitoring solution for the gateways and their customers' terminals.

Challenge

We were asked to implement intelligent methods for traffic control and rate limiting based on SES top clients' requirements, in the form of a Real-time Traffic Manager (RTM) application. As a network provider, this was a critical missing piece of SES global solution in order to ensure its customers have the best possible service at all times, regardless of the weather conditions.

Solution

With a predictive maintenance and a machine learning approach, now is possible to anticipate the outage, and manage it with the strategy that is best suited for the particular scenario. The system was deployed to be used in real-world production scenarios ensuring high quality Internet connections to SES clients.

As a result, our partnership with SES was able to transform a complex idea into a real product through a very productive working environment by reaching clear sprint goals. The first production deployment took place after about 11 months of development and extensive testing with the hardware systems.

95% reduction

in outage prediction time

Discover our approach

The RTM was built in Python, using the Flask stack to provide a RESTful API that exposes configurations and real time data.

The app communicates with external services and uses networking protocols to talk to devices, from which it takes data used as input for the algorithms.

Real time metrics are collected and stored in a time series database, so we can later gain further insight into traffic patterns. For testing we made heavy use of the mock library to simulate responses from devices.

Finally we built customized dashboards to display the most relevant statistics needed for a customer or system operator.

Tryolabs' team was not only able to rapidly adapt to our own set of processes and culture, but to quickly grasp telecommunications concepts previously unknown to them.

They were responsible for the software layer of an important part of our solution, implementing agile methodologies, great technical skills, daily communication and even on-site visits to our offices.

Andre Christian, VP Product Design and Engineering | SES

Discover how the world’s leading enterprises partner with Tryolabs for building data-driven solutions that boost their KPIs.