The approach to predicting data center cooling performance in real time pioneered by Schneider Electric is to focus on airflow patterns rather than temperature predictions. Airflow patterns, after all, are the root cause of cooling success or failure whereas temperatures are merely a symptom. It is possible to have acceptable temperatures while the management of cooling airflow is out of control and inefficient. Additionally, while airflow patterns are dominated by characteristics of nearby racks and coolers, temperatures "float" up or down with the surrounding ambient temperature. Consequently, it is easier to develop real-time cooling-prediction algorithms which predict airflow pattern characteristics rather than predicting temperatures directly.
The primary metric developed for airflow-pattern prediction is the Capture Index (CI) which is defined as either the fraction of a rack's airflow which comes directly from local cooling sources (Cold Aisle CI) or the fraction of the rack's airflow which is captured by local sources (Hot Aisle CI).
Once airflow patterns are known, however, it is possible to estimate all of the temperatures of primary interest (rack and cooler inlet and outlet temperatures and a room ambient temperature).