Artificial Intelligence assisted Canary Testing of Cloud Native RAN in a mobile telecom system

Abstract: Recent advancement in cloud-native infrastructure has made most organizations to transition from a traditional infrastructure of separate, static physical systems to cloudenvironments running on virtualized resources. Undoubtedly telecommunication industry will largely be benefitted from cloud-native infrastructure. In the future,network applications in Radio Access Network (RAN) will be built on cloud-nativeprinciples denoted as CloudRAN. In CloudRAN, new versions of the network applications that are integrated or deployed need to be validated before release. Canary testing is a popular testing strategy where the new version is exposed to a small subset of users initially. The performance of the new version is then monitored and analyzed to test and decide the quality of the new version. Unlike 4G, the 5GCloudRAN for the public mobile broadband may consist of hundreds of clusters and thousands of different microservices. Traditional DevOps solutions cannot keep upwith 3Vs of big data i.e. the volume, velocity, and variety. Furthermore, performing the analysis manually during canary testing is an exhausting process. In this thesis work, the problem of automating the decision-making process in canary testing ofCloudRAN applications by monitoring and analyzing time-series metrics of existing production version against new canary version using artificial intelligence methods is addressed.