Acoustics Analysis of Air and Hybrid Cooled Data Center

Due to advancements in technology nodes, the processor power and heat density is ramping up exponentially to meet users' and computational demands. Air cooling has been a traditional approach for thermal management for systems in the data center where conditioned air is provided from data center mechanical systems to the cold aisles. Server fans are used to facilitate the airflow through IT system heatsinks to remove heat from microprocessors. As a result of increased microprocessor energy consumption and the end of Moore's Law, the thermal margin is shrinking to maintain the required junction temperature limits. Data center operators are compelled to use lower inlet temperatures and higher flow rates to meet the thermal requirements and increase the capacity of heat exchangers. Increased flow rates have introduced additional fans, increasing noise levels of operating data centers. Primarily data center noise sources are the servers' fans, network equipment, heat exchangers (CRAC, CRAH, Rear-door HX, In-row cooler, Sidecar HX, rack mount HX, etc.), and tiles. Due to increased noise levels, data center operators must be aware of and follow OSHA guidelines for hearing protection.This paper estimates the acoustics from server fans and heat exchangers by analytical method, simulation with ray tracing method, and validation to quantify the effect of noise in a standard data center environment with an HPC POD consisting of multiple servers, racks, and aisle containment. Generated noise is mapped with servers as noise sources and the background noises with experimental measurements. Server-level noise calculations are performed analytically, numerically, and validated experimentally prior to the case studies in the data center. The noise level is expressed in terms of the Sound Power Level (SWL) and Sound Pressure Level (SPL) for the source and measurements. The case studies in the data center are experimentally validated for pure air-cooled and hybrid (air and liquid) cooled environments separately when Liquid -to-Air (L2A) heat exchangers are placed and operated inside the data center. The baseline case is considered for the 100% air-cooled environment. Results are presented for varying the airflow requirement per server for the cooling demand of air-cooled vs. liquid-cooled components when there is a cooling transition from 100% air to partial air-liquid hybrid cooling. The outcomes of this paper will be highly beneficial in finding the limits for the operation of servers and heat exchangers (in terms of temperature, flow rate, and humidity) and establishing safety guidelines for noise reduction in air and hybrid-cooled data center environments.