Adaptive monitoring in multiservice systems

Abstract

Understanding of complex systems is a key problem in the IT industry. Such systems can be imagined as a network of micro-services communicating among themselves. After the initial setup the configuration of such system changes with time. It is hard to follow the changes and always know configuration of the system. It is hard to detect errors in its processes and more important it is difficult to evaluate in a real time whether the system still works as designed. Aim of this paper is to provide a solution to evaluate the quality of processes conducted by the system in real-time. We offer complete fully automated solution, which can be easily implemented. This paper also briefly covers the problem of data normalisation, streaming and batch processing which is important for model creation and further validations. This kind of topology is known as lambda architecture. Furthermore we provide solution for graphical representation of the system. The system can be envisioned as a directed graph with micro-services as its nodes and communication between micro-services as its edges. We strive to graphically represent the system as a directed graph and also represent errors our solution detected. Therefore our work covers back-propagation together with model persistence and changes tracking over time.