多核系统已经被广泛应用于各行各业,其稳定性和可靠性也越来越受到人们的关注.在分析了现有芯片可靠性模型的基础上,增加温度和工作负载对多核芯片可靠性的影响,建立了对多核系统的可靠性评估模型.该模型以处理器为依托,从指令到任务,逐层构建可靠性评估指标,以便于定量分析影响多核系统可靠性的因素.同时,以该模型为指导,设计了一种面向多核系统可靠性的任务调度方法,该方法通过对评估指标值的计算,选择评估指标值尽可能高的调度策略对多核系统中的任务进行调度,以减少由于芯片本身可靠性而导致的错误.通过模拟实验可以看出,该任务调度算法能有效减少系统的出错率45%左右,为系统的稳定运行提供了可靠的保证.
The stability and reliability of multicore systems have been gotten more attentions with the widely used in all kinds of fields.After analyzing the current reliability model for chips,a reliability evaluating model considering the temperature and workloads for the multiprocessor is proposed.Based on construction characteristic of the multiprocessor,the model hierarchically constructs the evaluating index from instruction level to the task level that enables to quantitatively evaluate the reliability of the multiprocessor.To check the effectiveness of this model,a stability-oriented task scheduling algorithm is designed to enhance the stability of multicore system.It calculates the evaluating index for each scheduling scheme and chooses the scheduling scheme with highest evaluating index to obtain the low error rate caused by multicore.According to the results of simulation experiment,we can see that this model and method are effective that have the potential to reduce 45% of the error in multicore system.