Monolith versus Microservices: Scalability in Financial Enterprise Applications

Abstract

Financial enterprises face sustained pressure to scale transaction-processing systems that were, in many cases, originally built as monolithic applications, and the choice between continuing to scale a monolith and decomposing it into microservices is among the most consequential architectural decisions such an organization makes. This review synthesizes the peer-reviewed literature on monolithic and microservice architectures to examine what is actually known about their comparative scalability, with particular attention to the properties of financial enterprise applications specifically: strict transactional consistency requirements, regulatory separation-of-duties obligations, and latency-sensitive workloads. The review draws on empirical comparisons of monolithic, microservice, and serverless deployment costs and performance (Villamizar et al., 2017), systematic mapping of the microservices literature (Pahl & Jamshidi, 2016), empirical studies of microservice design-pattern trade-offs (Vale, Correia, Guerra, Rosa, Fritzsch, & Bogner, 2022), workload characterization specific to microservice systems (Ueda, Nakaike, & Ohara, 2016), the architectural trajectory of microservices as a style (Dragoni et al., 2017), the foundational distributed-systems constraints financial transaction processing operates under (Gilbert & Lynch, 2002; Helland, 2017), the distributed data-processing engines microservice-based financial systems commonly rely on for analytics and stream processing (Zaharia et al., 2016; Carbone et al., 2017), and the separation-of-duties enforcement financial compliance specifically requires (Botha & Eloff, 2001). The review finds that microservices offer independently scalable, fault-isolated components well suited to the heterogeneous load patterns financial applications exhibit, but that this benefit is not unconditional: it depends on a financial system's specific consistency requirements, operational maturity, and the genuine, well-documented cost of distributed-systems complexity that a monolith does not incur. No quantitative figures in this review represent newly collected empirical data; illustrative comparative figures, explicitly labeled as such, are included to synthesize patterns the cited literature reports.