Experimental validation of a low-cost microcontroller-based rack-level thermal control prototype

Abstract

The rapid growth of data centers (DCs), driven by digital transformation and the increasing adoption of artificial intelligence (AI), has intensified challenges related to thermal management and operational reliability. Cooling systems account for a substantial portion of total energy consumption and often show limited effectiveness in mitigating localized hotspots and dynamic temperature variations in high-density server environments. This study presents the development and experimental validation of a low-cost, microcontroller-based (MCU-based) localized thermal control system. The proposed architecture integrates a temperature sensor, an Arduino-based control unit, pulse-width modulation (PWM) driven fan actuation, and Ethernet communication for remote monitoring. The system was implemented in a standard 19-inch rack under controlled laboratory conditions using a simulated thermal load. Experimental results, based on the average of five independent tests, demonstrated that combined operation of the prototype with rack ventilation reduced the cooling time from 45 °C to 40 °C to 50 ± 2 s, compared to approximately 5 minutes with rack ventilation alone and more than 12 minutes under natural convection. The corresponding cooling rates were 0.10 °C/s, 0.015 °C/s, and 0.007 °C/s. These results indicate that simple, distributed thermal control strategies can effectively mitigate localized overheating and support rack-level thermal stability in data center microenvironments.