Version-based Microservice Analysis, Monitoring and Visualization

• Main Authors: Liu, I-Hsiu, 劉邑修
• Other Authors: Ma, Shang-Pin
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/d6crjs

碩士 === 國立臺灣海洋大學 === 資訊工程學系 === 107 === Microservice Architecture (MSA) is an emerging software system architecture. MSA uses microservices as system components, which are only responsible for specific tasks, have independent lifecycles, and collaborate with each other through lightweight communication mechanisms. Notably, the nature of good modularity for microservices yields better fault tolerance and scalability. However, since microservices need to communicate with each other to construct full functionality and multiple versions for each microservice running at the same time are quite common, the dependency relationship is very likely to be intertwined. Semantic Versioning is one of the solutions to the problem by limiting the configuration and growth of service version numbers through rules. Although semantic Versioning can suppress the growth of MSA's complexity, the interactions of various elements (such as services, endpoints, versions, etc.) are still complex and difficult to be managed. To effectively reduce the complexity of microservice management, we propose a novel approach, referred to as VMAMV (Version-based Microservice Analysis, Monitoring and Visualization), to analyze and visually display the microservice dependency relationships with semantic versions and to instantly reflect the states of microservices. VMAMV provides various types of search services based on the dependency graph, including searching for Consumers, Providers, Strong Upper Dependency Chains, Strong Lower Dependency Chains, Weak Upper Dependency Chains, and Weak Lower Dependency Chains. Meanwhile, VMAMV can automatically analyze the potential problems of system design, combine monitoring service metrics, methods of SPC (statistical process control) and user-defined thresholds to detect service anomalies, immediately notify users of the problems before or at the beginning of risky events. VMAMV detectable problems include Heavy Upper or Lower Dependency, Null Dependency, Old Patch Version is Still Running, Service High Failure Rate, Service High Response Duration, Service Frequent Internal Errors, Service High Response Duration, and Low-Usage Service Version.