工业物联网中大规模受损边缘计算网络修复机制

Recovery mechanism of large-scale damaged edge computing network in industrial Internet of things

针对工业物联网中边缘计算网络与其余子网的相互依赖特性所导致的网络大规模级联故障问题,考虑到网络修复初期的资源有限性,提出了一种联合考量计算需求与修复开销的网络修复机制。考虑到受损网络结构(拓扑关系与链路容量)和动态特征(边缘计算节点计算需求),基于节点计算量守恒定理构建了链路修复策略集与网络计算迁移的联合分析框架。基于Benders分解算法,将原NP-hard问题转化为相互依赖的主问题与子问题,通过割平面的不断逼近,实现对原问题最优解在多项式时间内的高效探索。结合局部分支法,进一步保障Benders分解算法的上界在迭代过程中的非增特性,加快算法收敛速度。仿真结果表明,所提算法的系统总开销性能优于传统基于拓扑结构的修复算法,并且可以在多场景下保持其性能优势。

Given the limited resources at early stages for recovery,a failure recovery mechanism of the edge computing network considering both computational demands and repair costs was proposed,which intends to tackle the problem of the high probability of large-scale cascading failure caused by the interdependence between the edge computing network and other subnetworks in industrial Internet of things(IIoT).Considering the network structure(topology and link capacity)and network dynamics(computational demands),a joint link recovery selection and computation migration optimization problem was formulated under the conservation of node computing requirements.By leveraging the Benders decomposition algorithm,the NP-hard problem was transformed into a main problem and a sub-problem,which were interdependent and could be solved in polynomial time through the approximation of cutting planes.A local branching method was further introduced to guarantee the non-increasing nature of the Benders upper bound,thus accelerating the convergence of Benders decomposition.Simulation results demonstrate that the proposed algorithm outperforms the conventional topology-based recovery algorithm in system utility,and can perform well in multiple scenarios.