A data center model for testing control and optimization algorithms

This paper describes a dynamical nonlinear data center model. It is a principle-based continuous time model, aimed at testing and evaluating control and optimization algorithms. The model is based on a module in data center used by RISE, Research Institute of Sweden. The data center modelled is a small scale slab floor data center. It contains four computer-room-air-handling units, two rows of five server racks, where each server rack has 18 servers. The model includes, as inputs, IT load of each server, airflow and temperature of each cooling unit, as outputs, the output temperature of the server. This model is a powerful tool simulating data center control-system prototype used to develop and test control and optimization strategies.