ARIMA time Series Model vs. K-Means Clustering for Cloud Workloads Performance
Vishnu Kumar Mishra, Megha Mishra, Sunil U. Tekale, T Naga Praveena, Rachakonda Venkatesh, Bhupesh Kumar Dewangan · 2023
To develop a thriving ability plan and protect the observable of Internet service providers of cloud environment, the increased heterogeneity brought on by various Cloud workloads, such as Business Data analytics, Big Data, IoT and calls for exact sensors. Although K-Means is an easy and quick clustering technique, it might not fully account for the heterogeneity present in Cloud platform workloads. The Multiple patterns can be found using ARIMA time series Models, which is trained for data prediction on cloud be combined into cohesive, homogenous components that closely resemble the data set’s actual patterns. In order to assess the cluster suggest the two different approaches for heterogeneity in resource utilization of Cloud infrastructure, this study compares ARIMA time series Model and K- Means. Clusters generated using K-Means yield significantly abstracted information, according to Bitbrains’ experiments using Google cluster trace and business-critical workloads. A more accurate clustering with discrete usage boundaries is provided by the ARIMA time series Model. Even though the ARIMA time series Model takes longer to compute than K-Means, it can be employed when a finer-grained characterization and study of the workload is needed.