Design and implementation of multi-layer tracing for microservices

Microservice architecture is an architectural design of the cloud services, that is already used in the major service providers, such as Amazon and Netflix. Using this pattern, a conventional monolithic service is divided into multiple microservices, then each microservice runs in the container that is light-weight virtualized technology. However, since a single system is provided by many microservices, it is difficult to understand the dependencies among microservices and to monitor traffic latency and error rates. When an abnormality occurs in a system with insufficient monitoring, it takes time to identify the cause of the abnormality and the service that is the bottleneck. To solve these problems, distributed tracing technology is used. This facilitates the analysis of increasingly complex applications. However, existing tracing technologies are unable to determine whether a large delay is actually occurring at the application layer, in the queue of the communication protocol stack, at the node driver level, or in the relay network. Therefore, in order to be able to identify system failures in more detail, it is desirable to be able to trace requests not only at the conventional application layer, but also at other layers. In this study, we propose multi-layer tracing, which enables analysis including transport and network layers