Modeling of data center airflow and heat transfer: State of the art and future trends

Abstract

An assessment of the current thermal modeling methodologies for data centers is presented, with focus on the use of computational fluid dynamics and heat transfer as analysis tools, and model validation. Future trends in reduced or compact modeling of data center airflow and heat transfer are presented to serve as an overview of integrating rack-level compact models into full-scale facility level numerical computations. Compact models can be used to efficiently model data centers through varying model fidelity across length scales. Dynamic effects can be included to develop next-generation control schemes to maximize data center energy efficiency.

A study on the effect of ground surface boundary conditions in modelling shallow ground heat exchangers

“…As reported by Crawford et al [31] for a building integrated photovoltaic system with heat recovery unit, heat recovery systems have a dominant impact on reducing the payback period of the energy supplying system. Future predictions indicate that the heat recovery systems will play an important role in the energy sector as a consequence of the developments in material sciences and growing significance of environmental issues [32][33][34][35].…”

“…According to Schmidt etal. [32], a typical 0.5MW dater centre with 50 server racks can be operate at AFR of 1500CFM per rack, the temperature difference between cold and hot aisles is 10.5 o C. A high efficiency air-cooled system [33] involving variable speed fans servicing for 28 high power density racks were studied, the maximum AFR is 2400CFM, and maximum temperature difference is 40 o C when the cold aisles at 15 o C, the cooling energy demand is 10%-15% lower than air -cooling system with fixed speed fan. A chiller or cooling tower is normally used to remove the heat returned from hot aisles to external environment.…”

“…The core strategy of air flow management is to minimize bypass air, which returns to the cooling units without performing cooling and the resultant recirculation and mixing of cool and hot air increasing equipment intake temperatures. A number of parameters will affect the air flow efficiency, such as width of hot and cold aisles, room height, distance between server rack`s top and ceiling [35]. The airflow recirculation, created by the turbulent mixing of hot and cold airs over the racks, was identified as the major cooling problem in air flow management designs.…”

“…Besides, through control the temperature at a 60 o C -85 o C, the application of water-cooled system in server or rack level are able to increasing the processor`s performance by 33% [42]. Based on the statistic research of air-cooled cases, typical parameters of water-cooled system in DC was summarised by researchers [35] in table 3 below. For the DCs with extremely high heat density, especially over 1Kw/cm 2 , the two-phase cooled system is considered instead of water-cooled system.…”

“…For the DCs with extremely high heat density, especially over 1Kw/cm 2 , the two-phase cooled system is considered instead of water-cooled system. The application of two-phase cooled system can be found from the surface heat flow between 0.7 Kw/cm 2 to 27Kw/cm 2 [35] due to its high heat transfer efficiency from internal evaporation and huge effective heat exchange area given by micro-channel heat pipe. Compare to water-cooled system, benefited from high efficiency of heat absorption by coolant evaporation, the mass flow rate, facility size and pumping power demand for a typical two-phase cooling system is separately 4 times, 2 times and 10 times less than the water-cooled system for same server [43].…”