A Cost-Capacity Analysis for Assessing the Efficiency of Heterogeneous Computing Assets in an Enterprise Cloud

Cloud providers and organizations with a large IT infrastructure manage evolving sets of hardware resources that are subject to continual change. As existing computing assets age, newer, more capable and more efficient ones are generally acquired. Significant variability of hardware components leads to inefficient use of computing assets within the organization. We claim that only a detailed understanding of the whole infrastructure will lead to significant optimizations and savings. In this paper we report results on a dataset of 1,171 assets from two different data centers, on which we present a thorough analysis of how the costs of running a computing asset are related to its resource capacity (i.e., CPU and RAM). This analysis is formalized in a cost model that could be used by organizations to make an optimal decision with regards to which computing assets should migrate their workload (i.e. should be disconnected or discarded) and which ones should receive such workload.