Modeling Impact of Power- and Thermal-Aware Fans Management on Data Center Energy Consumption

Abstract

In this paper we study the power usage and thermal management of micro servers to analyze their impact on the overall data center energy consumption. We propose thermal models of micro servers based on analytical approach tuned with parameters derived from empirical tests. We demonstrate how fan management configuration affects the energy consumption of servers and the whole data center. We also apply the proposed model to predict temperature changes in a short time ahead and take advantage of these predictions to improve fan management. We show why PUE is not sufficient or can be even misleading in minimizing data center energy consumption. To mitigate this issue, we propose metrics that can be used to reflect correctly fans management impact on the overall energy consumption.

Toward a Systematic Survey for Carbon Neutral Data Centers

“…The fans circulate air, carrying heat away from the server or rack. All heat generated by the server can be removed as the fan speed increases, and the server temperature will not rise [103]. As a result, the leakage power which monotonically increases with the server temperature will not increase, leading to less server power consumption.…”

“…In addition to server energy consumption, fans also consumes a considerable portion of energy, sometimes up to 51% in some server configuration [166]. Hence, quit a few researchers investigated the fan management problem so that either the energy consumption of fans or the total server rack can be minimized [103], [158], [159], and the approaches for optimizing server fan speed are summarized in Table IX. In [158], Wang et al investigated optimal fan speed control for a blade server.…”

Research advances on AI-powered thermal management for data centers

“…Piatek et al [31] adopted the resource efficient computing system (RECS), which has an embedded power box. In CoolEmAll, Cupertino et al [24] targeted the RECS version 2.0, which has a precision of 1 W as power meters that may not be enough for some usages.…”

“…Apart from the aforementioned models, some prediction tools were also introduced. For instance, Piatek et al [31] introduced microservers and thermal models based on an analytical approach tuned with parameters procured from empirical tests, which predicts temperature changes in a short period and is used to improve fan management. Cupertino et al [24] provided a description paper of the CoolEmAll project, which builds a toolkit that includes a data center monitoring, simulation, and visualization software.…”

Reuse of Data Center Waste Heat in Nearby Neighborhoods: A Neural Networks-Based Prediction Model

“…The models define the temperature of the processing unit, as well as the changes in the temperature at the server's outlet. Extension to these models, including the power leakage phenomena is described in [21]. Moreover, in [20] also the power models for the whole data center are proposed, together with the corresponding cooling models [22].…”