Abstract: Publication date: Available online 13 August 2018Source: Advances in Heat TransferAuthor(s): Jayati Athavale, Minami Yoda, Yogendra Joshi Thermal management of data centers remains a challenge because of their ever-increasing power densities and shrinking server volumes. At present, a lack of effective dynamic control of global provisioning and local distribution of cooling resources often results in excessive, and hence wasteful, cooling. These trends have motivated research on reliable and energy-efficient control frameworks for allocating cooling resources to meet thermal management requirements, while minimizing energy consumption and environmental impact. Key components of a dynamic framework are computationally efficient and accurate models of airflow and thermal transport in data centers. However, the most accurate models, namely physics-based numerical simulations, are impractical for control systems applications, which require nearly real-time response. We first review compact thermal modeling approaches for data centers, which could be suitable for dynamic control and optimization. This is followed by an assessment of control strategies for data centers focusing on need-based dynamic provisioning of both local and global cooling resources. It should be noted that at present, energy savings, if any, are a secondary outcome, and not the primary goal, of these control strategies. Recent work has considered incorporating both cooling energy minimization and meeting thermal management criteria as explicit objectives within a holistic optimization framework. The energy-efficient operation of data centers then becomes a multi-objective optimization problem. A key component for such an optimization framework is an efficient cooling energy estimation model, typically based on thermodynamic modeling of the data center cooling equipment. This chapter ends with a review of cooling energy minimization studies in data centers, along with progress toward a novel genetic algorithm (GA)-based optimization framework developed at the Data Center Lab at the Georgia Institute of Technology.

Abstract: Publication date: Available online 13 August 2018Source: Advances in Heat TransferAuthor(s): J.M. Gorman, E.M. Sparrow, A. Ghosh, J.P. Abraham The method of numerical simulation has been used here to establish the heat transfer characteristics of a fluid jet impinging on a target surface. In particular, it was demonstrated that a fluid mechanic phenomenon designated as jet axis switching has a tremendous effect on both the magnitude and surface distribution of the impingement heat transfer coefficient. The neglect of this phenomenon, which has been common in the literature on jet impingement heat transfer, gives rise to significant inaccuracies in the predicted values of the heat transfer coefficient. As an essential prelude to the heat transfer analysis, the fluid mechanics of jet impingement were set forth in detail in order to document the jet-axis switching phenomenon. For this purpose, color contour and velocity vector diagrams are displayed to show the change of shape experienced by the jet as it passes through an unconfined space. The local heat transfer coefficient at all points of the impingement plate was determined. The highest values of the Nusselt number do not occur when the plate is nearest to the origin of the jet. Off-axis peaks of the local Nusselt number were found to exist at locations between z/b = 0 and 3 for respective impingement plate positions Xmax/b = 30 and 10. The numerical predictions compared favorably with experimental results from the literature.

Abstract: Publication date: Available online 30 July 2018Source: Advances in Heat TransferAuthor(s): V.R. Voller The characteristic length scale in heat conduction, a diffusion transport process, changes with the square root of time. As we consider composite structures with increasing complex architectures, however, we find that this classic result is not always the case. When multiscaled heterogeneity is present in the system of interest, our heat transfer calculations may reveal anomalous behaviors. Here, the effects of nonlocality and memory, induced by “fast-paths” or “holdups,” lead to space–time scaling relationships that differ from the square root of time. This article has two themes. The first demonstrates how anomalous transport is induced in heat transfer applications. The second provides the necessary analytical and numerical details of fractional calculus operators and illustrates how these constructs can be used to model anomalous heat transport.