Taxonomia de faltas para aplicações baseadas em microsserviços

Abstract

Microservice-based applications support an architectural style that allows the organization of distributed applications as a set of independent services to achieve scalability and maintainability. They have been widely used in the industry, however, this architecture can bring challenges regarding fault Tolerance, Fault Prevention, Fault Detection and Fault Handling activities. In the context of microservices, fault injection is more complex since those applications contains mostly assynchronous process. Therefore, a fault injection becomes even more expensive. Therefore, the above activities that require tasks such as fault injection and debugging become even more expensive. This way, the use of a fault taxonomy can be beneficial for performing software fault activities. Several studies in the literature have been handling the complexity of testing microservice-based applications such as investigating the root causes of microservices failures or providing frameworks that allow users injecting faults related to microservices. In this work, we conducted a systematic mapping study to catalog faults related to microservice-based applications to better support their development and testing and then create a fault taxonomy. The results and contributions of this study encompass various facets. Initially, a comprehensive catalog is presented, comprising 136 faults, of which 103 are distributed across 11 categories. Subsequently, a preliminary taxonomy is proposed, classifying 117 of these faults in accordance with Non-Functional Requirements (NFRs) and correlating them with specific characteristics of microservices architecture. The evolution of this taxonomy culminates in a final version that includes 106 faults, also categorized by NFRs and correlated with characteristics of microservices architecture. Additionally, the taxonomy’s classification scheme is presented, predicated on 6 NFRs (e.g., Performance, Security, Reliability), alongside the correlation of faults with 14 characteristics (e.g., Runtime, Data Storage) inherent to microservices architecture. Lastly, the work furnishes a comprehensive account of the taxonomy creation and evaluation process, affording insights into its applicability.