Cooling-aware node-level task allocation for next-generation green HPC systems

Abstract

Energy-efficiency is of primary interest in future HPC systems as their computational growth is limited by the supercomputer peak power consumption. A significant part of the power consumed by a supercomputer machine is caused by the cooling infrastructure. Todays thermal design is based on coarse grain models which consider the silicon die of the processing elements as an isothermal surface. Similarly feedback control loops uses the same assumption to modulate the cooling effort with the goal of reducing cooling cost and maintaining the silicon temperature in a safe working range. Recent processors development has brought into the market CPUs that integrate a large number of complex cores. Differently from massively parallel CPUs for which the area and power consumption of each core is very limited, the cores of these processors can consume tens of watts and thus, under heterogeneous workloads, creating significant thermal gradients. In this paper we first characterize the power and thermal characteristics of new serverclass Intel Xeon computing node based on Haswell v3 architecture considering both the computational and the cooling components. We show that these systems are characterized by significant ondie thermal gradients and that the current O.S. task allocation strategy is not capable of taking advantage of that, leading to max CPU temperature and extra cooling activity. To solve this issue we propose a novel task allocation strategy that reduces the cooling power while matching the HPC performance requirements.

A Scalable Framework for Online Power Modelling of High-Performance Computing Nodes in Production

“…To solve these issues, several works in the literature [4], [5], [6], [7], [8] propose to take advantage of proactive thermal and power management strategies. These strategies all rely on the availability of compact predictive power models, capable of predicting future power consumption and, even more importantly to build a clear understanding on the sensitivity of power consumption on workload parameters and hardware knobs that can be controlled at run time.…”

Modeling and Evaluation of Application-Aware Dynamic Thermal Control in HPC Nodes

“…Due to different materials present in the heat dissipation path, the thermal transient is multi-modal with time constants that vary from ms to tens of seconds. Beneventi et al [5] shows in an Intel based computing nodes with 36 physical cores, that the increased number of processors integrated in the same die generates significant thermal gradients and this thermal heterogeneity can be exploited by thermal/aware MPI task allocation to reduce the fan speed and power without impacting the application performance.…”

Robust Identification of Thermal Models for In-Production High-Performance-Computing Clusters With Machine Learning-Based Data Selection

“…To solve these issues, several works in the literature [4], [5], [6], [7], [8], [9], [10] propose to take advantage of proactive thermal and power management strategies. These strategies all rely on the availability of compact predictive power and thermal models, capable of predicting future power consumption and temperature of the system and, even more importantly, to build a clear understanding on the sensitivity of these on workload parameters and hardware knobs that can be controlled at run time.…”