Abstract :
Data centers and their cooling efficiency have become the subject of much attention in recent years. While some new high-efficiency data centers adopt designs that do not rely on cooling distribution via a raised floor plenum, many still do. In any case, the predominant cooling approach for the average data center is to use a raised floor. The increasing focus on cooling efficiency has resulted in considerable application of computational fluid dynamics (CFD) to design, assessment/trouble shooting and operation with a view to maximizing IT availability, IT capacity and cooling efficiency. Indeed, it is now widely used by people whose specialism is not CFD. Further, the variety of tools available to these nonspecialist users is also diverse, including simplified tools intended to provide fast modeling and simulation; general purpose tools that in principle can model almost anything; and data center-specific, full feature tools. Previous research has tended to focus on bulk effects in the data center. For example, Karki et al. (2007) looked at the flow rates through the tile in a small raised floor data center. And while other groups (Abdelmaksoud et al (2010, 2012)) have made studies, there is as yet little high quality data from which to decide how to represent complex perforated tiles in large data center scenarios where there are many tiles in a large floor plan. CFD tools apply a 3D grid to the model. They then solve each cell in the grid. So, the bigger and more complex the model, the greater the number of grid cells. Since grid cells require processor cycles and time to `solve´, it is generally agreed that the smaller the number of grid cells (without sacrificing engineering accuracy of results), the better. With all this in mind, this paper begins by presenting data from initial experimental studies that show how not only jet characteristics, but also the gross flow rate through a tile, may vary based on angle of attack of the airflow. Having established thi- , the paper then compares fully-detailed models with examples of simplification that reduce the grid cell simplified models to show that traditional (very simplistic) models are insufficient to capture the true behavior of the cooling airflow. However, this paper explains how considerable simplification of a detailed model is possible without jeopardizing the capture of the key characteristics of the emerging jet. Unlike previous papers, which have focused on gross effects and ignored the impact of cross-flow underneath the tile, this paper concludes that traditional simplifications using a single resistance, and accounting for the increased momentum leaving the tile created by the tile open area (sometimes referred to as a porous jump model), is insufficient to capture the performance of the tile in realistic flow scenarios.
Keywords :
air conditioning; computational fluid dynamics; computer centres; floors; jets; tiles; 3D grid; CFD tools; IT availability; IT capacity; bulk effects; computational fluid dynamics; cooling distribution; cooling efficiency; data center-specific tools; grid cells; gross flow rate; high-efficiency data centers; jet characteristics; perforated tile modeling; porous jump model; processor cycles; raised floor data center; raised floor plenum; Atmospheric modeling; Blades; Computational fluid dynamics; Cooling; Data models; Floors; Tiles; Computational fluid dynamics; IT availability; IT capacity; cooling efficiency; cooling path management; cross-flow; data center; floor grille; floor tile; raised floor; simplification;