Seamless energy-aware hybrid cloud

This paper presents a novel approach to sustainable computing targeting cloud-native platforms, a paradigm to develop and deploy scalable applications in cloud computing environments based on technologies such as containers and microservices. Here, services are allocated according to clean energy availability, as opposed to traditional sustainable computing approaches based on computational efficiency. This proposal is based on a preexisting hybrid cloud architecture, emphasizing minimal alterations to the original platform —a feature we term seamless. The main goal is to assess the feasibility of our proposal through a specific implementation deployed in a real production environment, thus extending beyond typical academic pilots in controlled environments. The setup includes a medium-size photovoltaic system, a heterogeneous cloud infrastructure, with services allocated in both public and private computing hardware, and a large educational cloud-native platform running on top of that. Based on this, we obtain a mapping of 11 W per virtual instance, representing the power consumption of the underlying physical hardware. This value may serve as a basis for extrapolating energy savings in analogous cloud-based deployments. In addition, building on this experience, we pose whether the management of a hybrid cloud architecture based on sustainable criteria for service allocation can be transparent —without noticeable underlying modifications— to the different stakeholders involved in their respective operations. To answer this, we embrace different perspectives beyond purely technical aspects, from the platform's end user to the cloud infrastructure managers. Moreover, we share our experience in terms of internal acceptance and institutional positioning, focusing on sustainability awareness. • An hybrid cloud architecture able to seamlessly move workloads is implemented. • The architecture is deployed over a real production scenario. • Workloads are moved without noticeable changes for end-users. • The architecture behavior does not significantly affect users. • Technical staff shows good acceptance regarding the architecture's deployment.