Workload and user experience-aware dynamic reliability management in multicore processors

Abstract

Reliability is a major concern for nanoscale CMOS circuits. Degradation phenomena such as Electromigration, Negative Bias Temperature Instability, Time Dependent Dielectric Breakdown worsen with transistor scaling. Dynamic Reliability Management (DRM) techniques reduce reliability loss at runtime by constraining operating points, but they face the challenge of reducing user experience degradation while meeting a lifetime target. In this work we propose a sensor based hierarchical controller for multicore processor DRM, exploiting the major gap between the time scales of workload variations and reliability loss. We improve performance and user experience by locally relaxing reliability-induced operating point constraints, while meeting them over the large time windows relevant for reliability. With respect to the state-of-the-art, our solution guarantees timely execution of 100% of latency-critical applications, and have a 4% performance improvement over the whole lifetime.

A Linux-governor based Dynamic Reliability Manager for android mobile devices

“…This is achieved by the technique in [10], which is a workload-aware DRM technique for multiprocessors based on a two-level controller. This technique monitors system reliability on a long time scale and adapts operating conditions to workload quality requirements on a short time scale.…”

“…Dynamic Reliability Management (DRM) is a set of techniques trading off processor degradation and performance at runtime [10], [15], [18]. Reliability is periodically assessed, and processor operating conditions are controlled to limit the degradation source (i.e.…”

Combined on-line lifetime-energy optimization for asymmetric multicores

“…However, if the adopted policies do not take into account how they affect the system lifetime, the architecture reliability might be significantly affected. On the other hand, those solutions that focus only on lifetime improvement ( [11,15,14,18,6]) are typically characterized by high energy consumption. This work proposes an approach for designing systems with combined on-line lifetime/energy optimization by exploiting the architecture asymmetry.…”

Workload Uncertainty Characterization and Adaptive Frequency Scaling for Energy Minimization of Embedded Systems

“…Proactive energy management involves predicting these dynamic workloads apriori to determine the most appropriate frequency for every phase such that performance constraint is satisfied while minimizing the energy consumption [1]. Studies have been conducted recently to use machine learning to determine the minimum frequency through continuous feedback from the hardware performance monitoring unit (PMU) [2]- [10]. These approaches suffer from the following limitations.…”

“…This classifier is queried at run-time for a given application to predict the workload, and select the frequency and thread packing such that performance is maximized under a given power cap. A workload aware approach is proposed in [10] based on control theoretic principles. Our proposed approach differs from these techniques by addressing the three limitations discussed in Section I.…”