EXPLORING THE IMPACT OF CRAH UNIT PARTIAL FAILURE IN HYBRID-COOLED DATA CENTERS

The escalating information technology (IT) loads in modern data centers (DCs) present formidable challenges for traditional room-conditioning systems. The heat dissipated from IT equipment has witnessed a significant surge due to the rapid development of data processing, retrieval, and storage, driven by changing technology trends and the growing demand for online services. This evolving landscape poses a substantial burden on air-cooling systems, pushing them to their limits, especially with the prevailing trend of rising power densities in microprocessors and the emergence of hot spots. Amidst these challenges, single-phase cold plate cooling is gaining traction as IT power densities experience a dramatic climb. However, the widespread adoption of this cooling method faces impediments such as the limited availability of chilled water supplies, constrained air distribution pathways, and the absence of elevated floors in many older DCs. In response to these limitations, liquid-to-air (L2A) cooling distribution units (CDUs) have emerged as an alternative method. By incorporating hybrid air and liquid cooling technologies, the industry aims to achieve precise, on-demand cooling through the utilization of various techniques. In the realm of hybrid cooling systems that integrate both air and liquid cooling technologies, a partial failure of the Computer Room Air Handlers (CRAH) introduces unique challenges. Such a failure has the potential to disrupt the delicate balance between air and liquid cooling components, leading to uneven heat dissipation. Moreover, the interdependence of liquid and air cooling in hybrid systems means that even a partial failure can trigger a domino effect, reducing the overall cooling efficiency of the system. This comprehensive study delves into the implications of partial failure in the CRAH unit within the high-power density racks of a hybrid-cooled DC. The investigation explores how this partial failure impacts various critical parameters, including cooling capacity (CC), supply air temperature (SAT), air flow rate, supply fluid temperature (SFT), and thermal testing vehicle (TTV) heater case temperatures. For the purposes of this study, two L2A in-row CDUs were utilized, with a combined total heat load of 129 kW supplied to three racks. The experimental setup is meticulously equipped with the necessary instruments for monitoring and assessing tests on both the liquid coolant and air sides. By addressing these issues, the research contributes valuable insights to the ongoing efforts to optimize data center cooling solutions in the face of evolving IT demands and technological advancements.