基于Rejuvenation和检查点技术的软件性能研究

Software Performance Analysis Based on Rejuvenation and Checkpoints

近年来 ,一种新兴的软件容错技术Rejuvenation得到了广泛研究。该文将这项技术与传统的检查点技术相结合 ,提出了一种保障系统可靠性的容错策略 ,利用流体随机Petri网 (FluidStochasticPetriNet,FSPN)这一模型描述语言 ,建立结合Reju venation和检查点技术的软件系统工作模型 ,并以执行这一策略后系统的丢包率和总损失作为性能指标。对模型的数值分析和仿真表明 ,这项技术较之单纯的检查点技术 ,可以大大降低丢包率 ,提高系统可靠性。最后 ,根据系统损失函数 。

Recently a novel fault-tolerant technique, namely Rejuvenation, has been extensively investigated. In this paper, a new fault-tolerant scheme, which combines Rejuvenation with traditional checkpoints, is presented to improve system reliability. System model under this scheme is built through FSPN (Fluid Stochastic Petri Net) with job lost probability and system cost as performance functions. Model analysis and simulation show the contribution of Rejuvenation to reduce job lost probability and increase system reliability, and the optimal Rejuvenation scheme is obtained to minimize system cost.