Researchers have developed new algorithms to enhance the reliability of data center networks (DCNs) by constructing fault-tolerant paths and maximum disjoint paths in augmented cube-based architectures. These breakthroughs address critical challenges in handling node failures and improving data transmission efficiency, which are vital as cloud computing and big data continue to drive DCN expansion. The findings were published on 15 June 2026 in Frontiers of Computer Science, a journal co-published by Higher Education Press and Springer Nature.

As DCNs grow in scale and complexity, even minor node or link failures can disrupt services, leading to data loss or downtime. Traditional routing strategies often struggle with slow fault recovery and suboptimal path lengths. To tackle this, a team led by Weibei Fan from Nanjing University of Posts and Telecommunications focused on the augmented cube-based DCN (AQDNₙ), a structure known for its scalability and high connectivity.

The team’s first innovation is a fault-tolerant path construction algorithm. It ensures communication between any two fault-free nodes even when up to 2n−2 nodes fail (where n is the dimension of AQDNₙ). By leveraging the recursive structure of augmented cubes and optimizing cross-edge connections, the algorithm efficiently bypasses faults.

Their second key contribution is the AQDN-DP algorithm, which constructs 2n−1 disjoint paths between any two nodes— the maximum possible number determined by AQDNₙ’s (2n−1)-connectivity. These paths avoid shared nodes or links, eliminating common-mode failure risks and enabling load balancing.

Comparative experiments against traditional BFS (Breadth-First Search) showed striking improvements: the fault-tolerant algorithm reduced average execution time by 16.37% and path length by 54.23%. AQDN-DP also demonstrated strong fault tolerance, with more parallel links supporting reliable transmission as n increases.


"We designed these algorithms to ensure data centers can handle failures gracefully while maintaining efficiency," said Weibei Fan. "This means more stable cloud services, faster data processing, and better resilience against unexpected disruptions."

Moving forward, the team plans to explore applications in larger-scale networks and integrate adaptive learning to optimize paths in real time.

DOI
10.1007/s11704-025-50487-w