Energy-Efficiency in Data Centers

Abstract

Moving from a still on-going work [1], this section reports the progress being developed towards energy-efficient operations and the integrated management of cyber and physical aspects of data centers. In particular, an integrated system composed by wired and wireless sensors is presented: it monitors power consumptions of the servers and environmental conditions, with the goal of achieving an overall reduction of data centers’ energy consumptions. As like as the EMMON system (Chap. 8), the architecture proposed is intended to be hierarchical, modular and flexible enough to achieve high temporal and spatial resolution of the sensor measurements, with negligible latencies of sensors’ reports to the data center management control station.Overall, the advantage of having fine-grained power and environmental measurements in this application scenario is twofold: (i) measuring the power consumption at the single server level has enormous benefits for the business logic of data centers’ owners, since they can offer services and billing to their customers based on the actual consumption, and (ii) although there are in literature models to predict heat-flows used in commercial Computer Room Air Cooling (CRAC) systems, those models often lack of spatial resolution, so the availability of micro-climate conditions would help improving those models as well as continue feeding them with real data will improve the reliability and accuracy of their forecasts. Fine-grained measurements are also the basis to provide different views of the system, each of them customized to different users. The proposed architecture allows to set the desired resolution of the readings upon user’s requests, for example to investigate some problems in a specific area (row, room or floor) of the data center building. Every single sensor can be configured by setting user defined alarms and trigger measurements reports adaptively, by changing or (re-)configuring specific thresholds at run-time.