Mathematical model of a microservice system based on a Bayesian network for the task of optimizing configuration taking into account cascading dependencies

Objective. The aim of the work is to develop a mathematical model of a microservice system to optimize the configuration of this system.

Method. Methods of graph theory, probability theory, and decision theory were used for modeling, namely probabilistic graphical models such as Bayesian networks that combine elements of all these theories.

Result. A mathematical model of a microservice system based on a Bayesian network is described. A modified inference algorithm for this network and a configuration optimization algorithm taking cascades into account are proposed. The time complexity of inferring the optimal configuration is estimated. An experiment applying the proposed algorithms to a model system of 15 microservices was conducted, demonstrating that the configuration selected using the proposed model and algorithms has the lowest percentage of time the microservice system was in a state of service level agreement violation.

Conclusion. The model and algorithms will help in selecting the optimal configuration by taking into account cascading effects, discrete and continuous parameters. Further conclusions should be drawn by integrating the model with existing orchestration systems.

1. Newman S. Creation of microservices. 2nd ed. – St. Petersburg: Peter, 2023 – 624 p.

2. Kendall D. Stochastic processes occurring in queue theory and their analysis by the nested Markov chain method. Matematika. 1959; 3(6):97-112.

3. Tsypkin Ya.Z. Fundamentals of the theory of automatic systems. Moscow: Nauka, 1977. 560 p.

4. Abraham I., Alvisi L., Halpern J. Distributed computing and game Theory: Combining knowledge from two fields. ACM SIGACT News. 2011;42(2): 69-76.

5. Mao H., Alizadeh M., Menache I., Kandula S. Resource management through deep reinforcement learning. Proceedings of the 15th ACM Seminar on current topics in networks. 2016:50-56.

6. Taran V.N. Bayesian networks for modeling complex systems. Proceedings of 2017 IEEE 2nd International Conference on Control in Technical Systems, CTS St. Petersburg, 25–27 October 2017. St. Petersburg, 2017; 240-243. DOI 10.1109/CTSYS.2017.8109535. EDN XXUEQX.

7. Koller D., Friedman N. Probabilistic graphical models: principles and methods. Publishing House of the Massachusetts Institute of Technology, 2009;1266 p.

8. Pearl J. Probabilistic reasoning in intelligent systems: plausible conclusion networks. California: Morgan Kaufman, 1988.

9. Ilyin V.A., Poznyak E.G. Fundamentals of mathematical analysis. Part I., Moscow, 1998, pp. 248-251.

10. Amosov A.A., Dubinsky Yu.A., Kopchenova N.P. Computational Methods for Engineers, Moscow: Mir, 1998.

11. Kubernetes documentation. Horizontal module of automatic scaling [Electronic resource]. URL: https://kubernetes.io/docs/tasks/run-application/horizontal-pod-autoscale / (publication date 07/26/2025).

12. Xu B., Li H., Pang U. et al. Bayesian network approach to fault diagnosis of hydroelectric power plants. Energy Science, 2019;7:1669-1677. https ://doi.org/10.1002/ese3.383.

13. Guoqiang L., Zengru D., Ying F. Cascading failures in complex networks with a community structure. International Journal of Modern Physics C(IJMPC), World Scientific Publishing Co.Pte.Ltd.2014;25(5)1-10.

14. https://doi.org/10.1145/3190507. Tao Chen, Rami Bahsoon, and Xin Yao. 2018. A Survey and Taxonomy of Self-Aware and Self-Adaptive Cloud Autoscaling Systems. ACM Comput.Surv.51,3,Article 61.June 2018:40.