Control Systems and Prediction Methods for IT Cooling Performance in Containment

Background:

Containment is becoming a standard modality in data centers today due to its inherent energy efficiency advantages. Cold aisle containment, hot aisle containment, chimney, enclosed racks and rear door heat exchangers are different forms of segregation between cold and hot air streams. The containment industry tends to seal the contained space to address local hot spots. It also is common practice to tune down the cooling units’ blowers in order to increase the Power Utilization Efficiency (“PUE”) of the facility. The challenge for such systems, however, is the airflow mismatch between cooling units and IT (server, storage, switch) that occurs because the external temperature sensors are agnostic to the heating rates of the internal components, with the latter accelerating inside the server due to the airflow reduction. This can lead to thermal shocks that can even lead to failure.

Technology Overview:

The current invention addresses airflow mismatch by employing flow curves testing methods that describe the exact aerodynamic behavior of the IT. The passive flow curve method (PFC) describes the passive airflow behavior of the chassis while the IT is not on or where the IT does not have fans. The active flow curve method (AFC) collapses the IT internal resistance and effective fans into one. The resulting curve can be fit to server fan speed, and thus can be incorporated into facility-level control equipment that adjusts the airflow through the server based on the input pressure and fan speed.

Advantages:

• The percentage of airflow surplus or reduction is reported to the controller.
• The controller modulates the cooling airflow and the containment of artificial leakage.
• Under airflow reduction, the controller can modulate an increase in cooling airflow, open leakage paths to maintain reliable operation, and avoid CPU throttling.
• The controller can modulate smart louvers that are mounted at the IT outlets.