Control-Theoretic Thermal Balancing for Clusters

Free registration required

Executive Summary

Thermal management is critical for clusters because of the increasing power consumption of modern processors, compact server architectures and growing server density in data centers. Thermal balancing mitigates hot spots in a cluster through dynamic load distribution among servers. This paper presents the Control-theoretical Thermal Balancing (CTB) algorithms that employ feedback control loops to dynamically balance the temperature of different servers. CTB features optimal controllers designed based on a difference equation model that captures thermal dynamics of clusters. Stability analysis and simulation results demonstrate that CTB can provide effective temperature balancing when servers' power consumption and ambient temperature deviate from estimation or change significantly at run-time.

  • Format: PDF
  • Size: 259.34 KB