Anomalies in Microservice Software Architecture

This work discusses a shift in software architecture design trends, specifically thetransition towards a heterogeneous structure of software architecture. This shift hasstimulated the growing popularity of the microservices architectural model. Despitethe flexibility, scalability, and capacity for independent deployment of software builton the microservices model, such software products possess certain drawbacks. Akey problem is the risk of inefficient and unpredictable computational resourceconsumption, caused by the suboptimal design of interactions between services andsystem components. Consequently, classical methods for monitoring resource usageprove insufficiently effective. The paper proposes the use of dependent types toverify the potential existence of anomalous resource consumption, implemented atthe software architecture design stage.The microservices architecture model is examined through the lens of categorytheory: individual services are treated as objects, and their interaction protocols asmorphisms. The functional elements of services are modeled using tools fromdependent type theory and functional programming. The proposed model canaccount for both temporal costs and the resource intensity of performed operations.The concept of anomalous computational resource consumption is formalized basedon an adaptation of the amplitude criterion.To formulate recommendations for the structure of a microservices architecture, acriterion is considered that is based on the membership of the function describingcomputational resource consumption in a special class of functions. This proposed class of functions is characterized by the fact that functions describing dynamicprocesses do not lead to the emergence of anomalous phenomena.The key result of the research is a formalized approach that, at the softwarearchitecture design stage, allows for establishing guarantees for preventing theoccurrence of anomalous computational resource consumption and providesrecommendations to avoid anomalous resource usage without the need for constantmonitoring during operation.FundingsThis work was supported by the grant of the state program of the «Sirius» FederalTerritory «Scientific and technological development of the «Sirius» FederalTerritory» (Agreement №3-03 date 18.02.2025).