Keith A. Whisnant

Temperature aware task scheduling in MPSoCs

In deep submicron circuits, elevation in temperatures has brought new challenges in reliability, timing, performance, cooling costs and leakage power. Conventional thermal management techniques sacrifice performance to control the thermal behavior by slowing down or turning off the processors when a critical temperature threshold is exceeded. Moreover, studies have shown that in addition to high temperatures, temporal and spatial variations in temperature impact system reliability. In this work, we explore the benefits of thermally aware task scheduling for multiprocessor systems-on-a-chip (MPSoC). We design and evaluate OS-level dynamic scheduling policies with negligible performance overhead. We show that, using simple to implement policies that make decisions based on temperature measurements, better temporal and spatial thermal profiles can be achieved in comparison to state-of-art schedulers. We also enhance reactive strategies such as dynamic thread migration with our scheduling policies. This way, hot spots and temperature variations are decreased, and the performance cost is significantly reduced.

Static and Dynamic Temperature-Aware Scheduling for Multiprocessor SoCs

Thermal hot spots and high temperature gradients degrade reliability and performance, and increase cooling costs and leakage power. In this paper, we explore the benefits of temperature-aware task scheduling for multiprocessor system-on-a-chip (MPSoC). We evaluate our techniques using workload characteristics collected from a real system by Suns Continuous System Telemetry. We first solve the task scheduling problem statically using integer linear programming (ILP). The ILP solution is guaranteed to be optimal for the given assumptions for tasks. We formulate ILPs for minimizing energy, balancing energy, and reducing hot spots, and provide an extensive comparison of their thermal behavior against our technique. Our static solution can reduce the frequency of hot spots by 35%, spatial gradients by 85%, and thermal cycles by 61% in comparison to the ILP for minimizing energy. We then design dynamic scheduling policies at the OS-level with negligible performance overhead. Our adaptive dynamic policy reduces the frequency of high-magnitude thermal cycles and spatial gradients by around 50% and 90%, respectively, in comparison to state-of-the-art schedulers. Reactive thermal management strategies, such as thread migration, can be combined with our scheduling policy to further reduce hot spots, temperature variations, and the associated performance cost.

Temperature-aware MPSoC scheduling for reducing hot spots and gradients

Thermal hot spots and temperature gradients on the die need to be minimized to manufacture reliable systems while meeting energy and performance constraints. In this work, we solve the task scheduling problem for multiprocessor system-on-chips (MPSoCs) using Integer Linear Programming (ILP). The goal of our optimization is minimizing the hot spots and balancing the temperature distribution on the die for a known set of tasks. Under the given assumptions about task characteristics, the solution is optimal. We compare our technique against optimal scheduling methods for energy minimization, energy balancing, and hot spot minimization, and show that our technique achieves significantly better thermal profiles. We also extend our technique to handle workload variations at runtime.

Efficient Signal Selection for Nonlinear System-Based Models of Enterprise Servers

Modern computer systems are equipped with a significant number of hardware and software sensors from which time series telemetry data can be captured for analysis. One particularly interesting application of the time series data is proactive fault monitoring- the ability to identify leading indicators of failure before the failure actually occurs. Advanced pattern recognition approaches based on nonlinear system-based models are frequently used in proactive fault monitoring, whereby the complex interactions among multivariate signal behaviors are captured. For such approaches, a model is constructed in the training phase, during which the (nonlinear) correlations among the multiple input signals are learned. In the subsequent surveillance phase, the value of each signal is estimated as a function of the other signals. Significant deviations between the estimates and observed signals indicate a potential anomaly in the system under surveillance. Choosing an appropriate subset of signals to monitor largely has been an exercise in engineering judgment, rudimentary linear correlation analysis, and trial-and-error. This paper presents a genetic algorithm approach at signal selection that efficiently identifies a near-optimal model based upon multiple criteria