Proactive temperature balancing for low cost thermal management in MPSoCs

Abstract

Abstract-Designing thermal management strategies that reduce the impact of hot spots and on-die temperature variations at low performance cost is a very significant challenge for multiprocessor system-on-chips (MPSoCs). In this work, we present a proactive MPSoC thermal management approach, which predicts the future temperature and adjusts the job allocation on the MPSoC to minimize the impact of thermal hot spots and temperature variations without degrading performance. In addition, we implement and compare several reactive and proactive management strategies, and demonstrate that our proactive temperatureaware MPSoC job allocation technique is able to dramatically reduce the adverse effects of temperature at very low performance cost. We show experimental results using a simulator as well as an implementation on an UltraSPARC T1 system.

Power-Thermal Modeling and Control of Energy-Efficient Servers and Datacenters

“…Previous work [82] exploits a temperature forecast technique based on an auto-regressive moving average model. Another work proposes a novel technique that adapts the thermal management policy to the current workload characteristics [76], where the adaptation is done online exploiting information related to the workload history.…”

Online thermal control methods for multiprocessor systems

“…In several recent works, history information has been exploited to improve thermal management policies. Coskun et al [2008a] exploits a temperature forecast technique based on an auto regressive moving average model. Lee et al [2010] and Kang et al [2011] propose policies for 2D and 3D MPSoC using autoregressive moving average applied to performance counters to find the correlation between the applied voltage setting and the measured MPSoC temperature.…”

“…In particular, the thermal policy techniques we consider are the linear quadratic regulator [Kang et al 2011;Coskun et al 2008a;Zanini et al 2009b] (i.e., unconstrained MPC with horizon equal to infinity), the explicit/implicit model predictive control-based approach [Wang et al 2009;Zanini et al 2009a] (i.e., traditional MPC), the approximated explicit model predictive control policy [Zanini et al 2010a] (i.e., approximated MPC), and finally, the convex optimization-based approaches [Hanumaiah and Vrudhula 2012;Zanini et al 2010b] (i.e., joint workload and thermal profile prediction). This last technique is solved with a convex solver; however, it is an MPC as well with a linear objective function.…”

Modeling and dynamic management of 3D multicore systems with liquid cooling

“…Due to thermal time constants, after we tune any system parameter such as workload scheduling or power consumption via dynamic management, there is a delay before we Proactive workload allocation has previously been proposed in [8]. On each core, we use an ARMA (Autoregressive Moving Average) predictor to forecast temperature.…”

“…p and q represent the orders of the auto-regressive (AR) and the moving average (MA) parts of the model, respectively. ARMA prediction is highly accurate for temperature forecasting, and runtime adaptation methods can also be integrated with ARMA as discussed in [8].…”