Further exploration of a compact transient server model

Two, similar compact server models have recently been proposed for the purpose of capturing the effects of server thermal mass in transient data center thermal simulations. Both models predict server exhaust temperature from a known inlet temperature at any time provided the server properties of thermal capacitance (heat capacity) and thermal effectiveness are known. One model also includes a third server property, λ, representing the relative position of the server´s idealized heat source to its mass, theoretically necessary when the power or airflow inside the server changes with time. Preliminary published measurements found that λ may be close to zero and therefore negligible for practical scenarios. If λ is truly negligible under practical conditions, the compact model and experimental measurements may be simplified. Based on experimental measurements of three 2U server configurations, it is shown that λ is not required. This paper also discusses a refined experimental procedure and apparatus for measuring server heat capacity and server thermal effectiveness. Server thermal effectiveness is found to be in the range of 0.85-1 with the larger numbers corresponding to more densely-packed (e.g., circuit boards, power supplies, and hard drives installed) server configurations. Additionally, the dependence of server thermal effectiveness on server airflow rate diminishes with increasing server mass density. Finally, an empirical correlation between server mass density and thermal effectiveness is provided which can be used as a first-order estimation in the absence of more detailed server measurements.