Enhancing Polyglot Microservice-Oriented Architecture with Event-Driven Concepts

Abstract

Financial institutions increasingly adopt microservice architectures to sustain competitive advantage through technological innovation. Yet many rely on synchronous requestresponse patterns that create tight coupling, cascading failures, and suffers from scalability limitations—having an enormous impact, especially on financial operations that require near-real-time response times. This thesis investigates how event-driven patterns can be effectively integrated into existing polyglot microservice architectures while maintaining banking-specific requirements including auditing, regulatory compliance, and data protection. Through comprehensive theoretical analysis and practical implementation, the research examines the effectiveness of Apache Kafka as an event streaming platform for polyglot microservice and legacy system integration. Following the theoretical analysis, event-driven architectural patterns, particularly emphasising Change Data Capture, are demonstrated through an extensive case study that implements event-driven integration within a simulated banking environment. Theoretical analysis indicates that hybrid communication strategies–—combining synchronous user-facing APIs with asynchronous back-end processing—–offer robust solutions within the banking sector. Event-driven publish-subscribe approaches enable loose coupling, enhanced scalability, load buffering, and dynamic scaling while maintaining high availability and consistency under varying loads. Apache Kafka’s use of standardised serialisation formats for language-agnostic communication supports seamless polyglot workflows across different programming languages and frameworks. The practical implementation validates the theoretical concepts of event-driven patterns and hybrid communication strategies through a fully containerised streaming data processing pipeline built around financial transactions. The end-to-end system successfully demonstrates Change Data Capture using Debezium and PostgreSQL logical replication, stream processing with ksqlDB and Kafka Streams for transaction categorisation, and multi-tenant topic organisation for data isolation. The implementation demonstrates significant improvements in system resilience and scalability. This thesis provides valuable insights for financial institutions modernising their architectures. Demonstrating integration patterns and implementation strategies help to achieve the scalability, resilience, and responsiveness required in today’s competitive banking landscape.

Date of Award	2025
Original language	English
Supervisor	Bogdan Burlacu (Supervisor)