Research On Cooling Technology For High Heat Density Data Center

With the rapid growth of information industry, demand for high performance datacenter is increasing. Blade servers have lifted the capacity of a normal42U rack fromless than1kW in early years to recent10~20kW, rising IDC heat density up to1~3kW/m2,and nearly40%of data center operating cost is spent on air-conditioning.Reducing cooling cost is an urgent issue which affects management and development ofhigh heat density data centers. This paper characterizes the thermal behavior and thegoverning rules through the analysis on heat transfer ability loss of data centers fromfollowing aspectsThe essence of data center heat transfer. The central work is to remove the givenheat under available driven force (temperature difference between indoor and outdoor),with minimum temperature difference dissipation and cold source of high temperature.The temperature distribution of a real data center indicates that the mix of hot and coldair has a significant impact on heat transfer performance.The analysis method of data center heat transfer. Maximizing heat flux undergiven temperature difference by flow rate optimization is performed using boththermodynamic method (entropy generation and exergy loss) and entransy analysis(entranssy dissipation). The results show that entransy analysis is suitable for heattransfer optimization by minimizing heat transfer ability loss, which is represented byentransy dissipation caused by temperature difference driven heat conduct, mixing ofhot and cold air and unmatching flowrate in heat exchangers. The influence of inputwork and flowrate change in heat exchanger network on entransy dissipation is alsoanalysed with test data.Distributed cooling system. Based on minimum principle of entransy dissipation.distributed cooling solution is brought up to improve data center cooling performance.Inner cooled racks with multi separated heat pipes inside eliminate hot and cold airmixing by multi-stage heat removal, and cooling water loop comprised of serialconnected cooling tower and multi-chiller enables continuous distribution of freecooling capacity and mechanical cooling capacity according to the outdoor condition,making more free cooling available. Such cooling loop has large temperature differenceof supply and returen water, which improves multi-chiller performance. With more free cooling and higher chiller COP, distributed cooling can be more energy efficient.Engineering application. Through the air-conditioning retrofitting project of a highheat density data center in Beijing, the performance of two cooling solutions, distributedcooling and centralized cooling, are tested and compared after one year operation. Theresults show that after using distributed cooling, the data center’s annual average EER(Energy Efficiency Ratio) of air-conditioning system increases from2.6to5.7, and thedata center’s annual average PUE (Power Utilization Effectiveness) decreases from1.6to1.35, indicating the applicability of entransy analysis and distributed cooling solutionin data centers of high heat density. This retrofitting project enriches the experience forthe widely use of distributed cooling system in data centers in the future.