An efficient parallel approach for binary-state network reliability problems

Networks are ubiquitous in modern applications, with reliability being a paramount performance metric. While exact network reliability can be determined using implicit enumeration algorithms like depth-first search, breadth-first search, the universal generating function methodology, the binary decision diagram, and the binary addition tree algorithm (BAT), these methods are limited to small-scale networks. The recently introduced BAT algorithm offers a high-speed, flexible, and easily implementable exact solution for determining network reliability. Experimental results demonstrate BAT’s superiority over other implicit enumeration algorithms. To address the computational challenges of larger networks, we propose a multithreaded version of BAT (mBAT). By leveraging multi-core architectures, mBAT efficiently solves medium-scale network reliability problems, as confirmed by time complexity analysis and experiments on 20 benchmark instances using up to 12 CPU threads.