Cloud Workload Categorization Using Various Data Preprocessing and Clustering Techniques

Effectively managing cloud resources can be challenging due to the inter-dependencies of various cloud-hosted services. Workload categorization identifies and groups workloads with similar characteristics. Data center managers can make informed decisions on resource allocation, workload scheduling, and infrastructure maintenance, leading to better performance and reduced costs. However, since cloud workloads can be interpreted differently due to their characteristics, several well-founded categories can be concealed within the various data perspectives. This paper proposes a workload categorization approach to automate the categorization process of the scheduling workloads, utilizing different clustering and data preprocessing methods, evaluated using a cloud workload trace derived from Microsoft Azure. Our research highlights the importance of using advanced data preprocessing techniques and integrating them seamlessly into clustering methods to ensure precise workload segmentation.