Comprehensive Benchmarking Analysis of Auto Scaling Approaches in Cloud Native Streaming Pipelines During Flash Sales and Holiday Traffic Peaks

Abstract

This study compares the behavior of the various auto-scaling policies in cloud-native streaming pipelines under the presence of retail surges like flash sales and holiday traffic. The comparison is made between Vertical Scaling, Horizontal Scaling, Predictive Scaling, and Adaptive Scaling on the basis of the resource utilization, availability of the systems and performance. Statistical analyses, such as ANOVA and t-tests, demonstrate that there are substantial differences in models in the way they use resources. The results are that Predictive and Vertical Scaling have an effective use of resources and Adaptive Scaling has the opportunity to be flexible in case of dynamic traffic. The study offers practical implications on the choice of the best scaling model to achieve maximum efficiency in a cloud environment in times of heavy retailing.