Machine Learning and Mobile Device Connectivity Optimization

Machine Learning and Mobile Device Connectivity Optimization

You must have experienced this common universal experience, that your phone is hanging and not functioning at a critical time. For example, you are using your phone for navigation and at a critical moment like ‘which way to go’ your phone is hanging and showing no connection. In Europe companies like Deutsche Telekom are focused on eliminating such problems by machine learning network management and machine learning algorithm, based on dynamic cloud infrastructure, to make consumers mobile connectivity experience.

You can actually download the Deutsche Telekom CONNECT app, from the ios App Store and Google Play launched the previous year. So that consumers can optimize connectivity based on cost or performance.

mobilenet

If you want to avoid slow performance or above like situation, then choosing ‘best network’ setting, is best. Where the app, to provide better experience automatically switches to cellular network from the Wi-Fi hotspot.  

Data Analytics and Machine Learning

The architect responsible for bringing ‘CONNECT’ app backend infrastructure of Deutsche Telecom is Oliver Goldich and team. He stated the project of ‘CONNECT’ was started in 2015 and began with investigating connectivity solutions and controlling an emerging telecommunication standard called the Access Network Discovery and Selection Function (ANDSF). But this standard was not only old-fashioned but was inflexible too and for this project Goldich told, they needed a dynamic rule engine(for dynamic monitoring and decision-making) that could not support by protocol.

So to adopt advanced data analytics and machine learning, Goldich and team required new kind and they began sourcing the component to reach there. And they found Mesosphere DC/OS, which was already a proven platform and was best to complete the “CONNECT” project. On a DC/OS service layer, Goldich and team built a network speed test app and the user was allowed to do a simple speed check. And to validate their approach Deutsche Telekom collect and analyzed that user data, later on.

Selection of Cloud Provider

For speed test automation and improving user experience, Goldich, and the team started to create infrastructure requirements, as results looking positive. So for developing machine learning capabilities, scaling was necessary and scaling cloud capabilities needed to be added and was significant to do so without vendor lock-in, Goldich stated. 

Goldich noted complications arise with the cloud. As deploying to an on-premise data center, there existed a DevOps process, that wouldn’t have worked with machine learning stack, Goldich and team want to adopt.

For setting up DC/OS environments with new CI/CD capabilities, they choose Microsoft Azure. It was a good step choosing a cloud provider like that as if they choose to move providers in the future, they don’t need to re-architect the applications. With DC/OS applications can be moved to preferred infrastructure such as cloud, bare metal, or on-premise, along with less minimal effort.

Deutsche Telekom maintains a high rate of utilization by the use of elastic cloud resources, which is allowed by DC/OS and also eliminates the need to create virtual machine. The high rate of utilization currently averages more than 75% CPU utilization on their production cluster.

You can get access to all of the open-source data by DC/OS, that’s what Deutsche Telecom needed. And further Deutsche was able to perform data collection at scale and analysis of network speed tests in real-time.

ELK Stack

To supplement their data science Deutsche use “ELK Stack” which means, Elasticsearch, Logstash, and Kibana and for AI-powered application performance monitoring partnered with Instana, for getting full deep insight into the health of their technology services.

deep reinforcement learning

When everything was set in place, Deutsche Telekom launched CONNECT app, at the end of 2017 and to 156 million mobile customers, providing better connectivity experiences. To collect data on the hundreds of available hotspots, Deutsche Telekom performs an automated speed test.

Developers time to time do incremental improvements and DC/OS allows developers to do so for customers because machine learning models evolve continuously. The big goal here is to include predictive analytics, determining when certain cellular networks or hotspots are normally overcrowded and shifting network usage accordingly.but for now, CONNECT app is focused on Deutsche Telekom’s network of hotspots.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

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