Abstract

Event-driven architectures (EDA) have emerged as a foundational paradigm for modern high-throughput computing systems, enabling asynchronous, scalable, loosely coupled interactions among components. Analytical performance modeling of EDA provides a quantitative basis for understanding system behaviors under variable workloads, resource constraints, and design choices. This paper investigates analytical models that characterize throughput, latency, queueing behavior, and resource utilization in event-driven systems. We examine stochastic modeling techniques, including queueing theory, Markov chains, and fluid approximations, to establish performance bounds and predict behavior under extreme loads. Our analysis extends traditional methods by integrating system parameters such as event arrival distributions, processing heterogeneity, and event dependencies. The results illustrate trade-offs between responsiveness and scalability, identify bottlenecks in event processing pipelines, and quantify the effect of architectural decisions on overall performance. Case studies demonstrate applicability across distributed event streams, serverless platforms, and actor-based systems. The findings guide system designers in optimizing event dispatching policies and resource allocation strategies. This work contributes a rigorous methodological framework to support performance engineering in high-throughput event-driven environments.