Understanding Scheduler Workload on Non-Hyperscale Cloud Platform

Abstract

Until now, research in cloud computing has mainly focused on hyperscale platforms. However, mid-size platforms represent 20%-40% of the worldwide public cloud market, and their usage under the form of on-premise cloud solutions is increasing. This popularity requires a better understanding of mid-size platforms' workloads, in order to optimize the way schedulers allocate resources. In this work, we study the scheduled workload of a mid-size platform and show it has different patterns than that of a hyperscale cloud.