Modeling Performance of Microservices Systems with Growth Theory

• DOI: 10.1007/s10664-021-10088-0
• Publication Date: 2022-01-11T09:02:57Z
• Authors: Matteo Camilli; Barbara Russo
• Keywords: Microservices; Benchmark (surveying); Resilience; Architectural style

The microservices architectural style is gaining momentum in the IT industry. This does not guarantee that a target system can continuously meet acceptable performance levels. ability to study violations of requirements and eventually predict them would help practitioners tune techniques like dynamic load balancing or horizontal scaling achieve resilience property.The goal this work through time series analysis provide practical instruments detect resilient non-resilient possibly their behavior.We introduce new method based on growth theory model occurrences as stochastic process. We applied our an in-vitro e-commerce benchmark in-production real-world telecommunication system. interpreted resulting models characterize terms transient behavior.Our empirical evaluation shows that, most cases, non-linear S-shaped capture with high accuracy. bounded nature associated tell degradation limited thus microservice able come back level even under changes nominal number concurrent users. also cases where linear represent better description. These are exhibit constant unbounded over time. application methodology real identified additional profiles were observed experiments. show services react differently same solicitation. found when service it either decrease rate continuous manner repeated attempts (periodical not).We showed be successfully occurences systems. Furthermore, cost calibration heuristics, mathematical expression selected models, limited. discussed how shed some light trend engineers spot problematic operations issues. Thus, meaningful insights from have been derived behavior (non) operations.