Bridging the White and Gray Space Data Center Silos
Integrating Workload Clustering and Zonal Cooling to Maximize Compute Per Megawatt™
Most data centers operate at only 40–60% of their true potential. The remaining capacity isn’t missing—it’s stranded by misalignment in how workloads are organized and inefficiencies in how cooling resources are deployed. Unlocking this capacity means unlocking revenue.
This white paper details PADO’s integrated optimization framework, which uses Workload Clustering and Zone-Targeted Cooling to minimize stranded capacity by decoupling power constraints with compute constraints. By bridging the traditional silos between white space (predictive job packing) and gray space (zonal cooling), PADO maximizes Compute Per Megawatt™. Historically, the operation of the white and gray space has occurred in silos with separate entities optimizing to different goals and in the process stranding capacity. The PADO approach, as described in this paper, connects the white and the gray, creating transparency in the white space to eliminate operational silos, thus increasing throughput via workload clustering with HVAC zone targeting.
Workload Clustering: Understanding specific job demands is essential for optimization. Different workloads—from steady AI training to bursty inference—impose unique demands on power and infrastructure. Clustering these jobs by a broader set of attributes (thermal profile, sovereignty, priority, etc.) allows for more precise capacity planning and cooling.
Zone-Targeted Cooling: This strategy complements white space optimization by matching cooling delivery to a zone's specific thermal needs. This precision allows for increased temperature setpoints in cooler zones while maintaining safety in high-density areas through variable control.
- 15-25% improvement in effective utilization
- 8-15% improvement in PUE
- 20-40% improvement in throughput density