Performance variability and workload bursts in asynchronous backend systems can lead to queue buildup, increased processing latency, and reduced responsiveness, making effective scaling strategies essential. Traditional resource allocation approaches often rely on static configurations that adopt poorly to changing conditions. This thesis investigates the performance implications of scaling strategies in an asynchronous queue-driven backend system, focusing on throughput, latency, and queue stability under varying workloads.

A discrete-event simulation (DES) framework is developed to model the behavior of an asynchronous file-processing pipeline inspired by a real-world system operated by ITS at Umeå University. The framework abstracts and models core system characteristics, including workload generation, queue-based task processing, worker execution, and scaling mechanisms, while intentionally excluding implementation specific infrastructure complexities.The study evaluates horizontal and vertical scaling strategies using both static and dynamic approaches under low, medium, and burst workload conditions. To ensure reproducibility and reduce the influence of stochastic variation, the study repeats experiments using multiple random seeds.

The results show that scaling strategies significantly influence system behavior under increasing workload pressure. Static and dynamic approaches both improve system performance compared to the baseline configuration, though their effects differ in responsiveness, queue dynamics, and resource utilization. Horizontal scaling improved parallel processing and reduced backlog formation under burst conditions, while vertical scaling improves processing efficiency by increasing worker capacity. Dynamic scaling approaches further demonstrate advantages in adapting system capacity to changing workload conditions while maintaining system stability.

The findings contribute a structured comparative evaluation of scaling strategies in asynchronous queue-driven systems and provide practical insights into how scaling decisions affect performance and stability. The proposed simulation framework also provides a reusable environment for evaluating future scaling policies and supporting data-driven infrastructure decisions in ITS and similar backend systems.