Distributed resource management in data center with temperature constraint

With the ever-increasing power density generating an excessive amount of heat that potentially induces server damage and outages, data center operation must be judiciously optimized to prevent server overheating. Moreover, data center is subject to various peak power constraints (such as peak server power) that have to be satisfied at all times for reliability concerns. In this paper, we propose a novel resource management algorithm, called DREAM (Distributed REsource mAnagement with teMperature constraint), to optimally control the server capacity provisioning (via power adjustment) and load distribution for minimizing the data center operational cost while satisfying the IT peak power and maximum server inlet temperature constraints. By using DREAM, each server can autonomously adjust its discrete processing speed (and hence, power consumption, too) and optimally decide the amount of workloads to process, in order to minimize the total cost that incorporates both power consumption and service delay. We rigorously prove that DREAM can yield the minimum cost with an arbitrarily high probability while satisfying the peak power and inlet temperature constraints. To complement the analysis, we perform a simulation study and show that DREAM can significantly reduce the cost compared to the optimal temperature-unaware algorithm (by up to 23%) and equal load distribution (by up to 69%).