Analysis and Optimization of MPSoC Reliability

Abstract

Advancements in technology enable integration of multiple devices on a single core, resulting in increased on chip power and temperature densities. Higher temperatures, in turn, present a significant challenge for reliability. In this work we propose a comprehensive framework for analyzing reliability of multi-core systems, considering permanent faults. We show that aggressive power management can have an impact on reliability due to temperature cycling. Our cycle-accurate simulation methodology shows fine-grained variations of device failure rates over short time scales, thus enabling workload analysis and scheduling to control the reliability impact. On the other hand, the statistical reliability simulator and optimizer give a view into the long time horizon reliability analysis-over system lifetime, and help us optimize a power management policy under reliability and performance constraints. We show that our optimization strategy can achieve large power savings while still meeting the reliability and performance constraints.

On Task Allocation and Scheduling for Lifetime Extension of Platform-Based MPSoC Designs

“…The above models target a single-core processor's lifetime reliability. Coskun et al [25] proposed a simulation methodology to evaluate the lifetime reliability of multicore systems, and used it to optimize the system's power management policy. For the sake of simplicity, most of the above models assumed exponential failure distributions and thus cannot capture the processors' accumulated aging effects.…”

Lifetime reliability-aware task allocation and scheduling for MPSoC platforms

“…The above models focus on analyzing single-core processor's lifetime reliability. Coskun et al [8] introduced two analytical frameworks for the lifetime reliability of multicore systems: a cycle-accurate simulation methodology and a statistical one, assuming uniform device density. Huang and Xu [15] proposed to model the lifetime reliability of homogeneous manycore systems using a load-sharing nonrepairable k-out-of-n:G system with general failure distributions for embedded cores.…”

Energy-aware Fault-tolerant Scheduling for Hard Real-time Systems

“…Consequently, we can complete the system reliability calculation. For a simple distribution, such as Exponential Distribution, sum-of-failure-rate (SOFR) can be used to get the system MTTF [113,38] while for much complicated distributions, e.g Weibull and Lognormal, Monte Carlo simulation needs to be conducted [126].…”