Cloud Computing Systems: Performance Trade-Offs In Scalable Architectures

Abstract

Cloud computing systems play a critical role in supporting scalable applications in domains such as e-commerce, where highly dynamic user demand requires efficient resource allocation. Auto-scaling mechanisms enable cloud platforms to dynamically adjust resources in response to workload variations while maintaining Service Level Agreement (SLA) requirements. However, achieving an optimal balance between system performance and operational cost remains challenging due to complex interactions among workload patterns, resource configurations, and scaling strategies. This research proposes an optimization based on a Bear Smell Search Feed Forward Neural Network (BSS-FFNN) model, where the BSS algorithm is employed to explore the search space and identify optimal resource configuration parameters, while the FFNN is utilized to learn workload patterns and predict system performance for informed decision-making in resource management. An e-commerce workload dataset comprising 11,000 samples is utilized for evaluation. Data preprocessing includes Min-Max normalization and outlier removal based Interquartile Range (IQR) method to ensure data consistency. Feature extraction is performed using Principal Component Analysis (PCA) to identify significant workload characteristics. The proposed method identifies optimal configurations that balance response time, throughput, and cost while ensuring SLA compliance. Experimental results demonstrate a response time of 36ms, a throughput of 600 req/ms, Central processing unit (CPU) utilization of 49 %, and memory utilization of 35 %, and the implementation is carried out in Python using deep learning (DL) libraries. In conclusion, the BSS-FFNN-based method provides a scalable and efficient solution for optimizing performance–cost trade-offs in cloud-based e-commerce environments.