Traffic analysis of a service mesh

Abstract

In this thesis we will look into the configuration of microservices. Over the years the software landscape has evolved from monolithic applications to microservice architectures. For these microservice architectures that configuration has become increasingly important. As the microservice architectures grew larger the amount of configuration has also increased which causes the developers to not fully grasp the mesh they are developing. This service mesh configuration allows the developer to automatically scale the application when demand increases. This configuration can also handle cases where a part of the network becomes unavailable or unreliable. By not exactly knowing how the network functions and what a small change can cause further down the network, errors can easily be introduced. We are investigating a way to give the developer more insight into the consequences of his changes. In a world where the software landscape is becoming increasingly important and large this can help developers to seriously improve the quality of their configurations. We looked into data from the configuration files for the service mesh and the logging that is produced by the service mesh. We found that with the information it is possible to create realistic simulations. Under the right circumstances it is even possible to make good predictions about the traffic and load within a service mesh after a change.