Fault tolerance assessment of the data center network DPCell based on g-good-neighbor conditions

Abstract Data center networks (DCNs) provide critical data storage and computing services for cloud computing. The continuous increase in demand for cloud computing has led to a surge in data volume, necessitating the continual expansion of DCNs. However, this expansion also heightens the risk of device failures. Therefore, it is particularly important to study the fault tolerance of DCNs, which refers to their ability to ensure reliable communication even in the presence of device failures. Among DCNs constructed using dual-port servers, DPCell achieves higher scalability and bisection width while maintaining a smaller diameter. This paper assesses the fault tolerance of DPCell using two metrics: connectivity and diagnosability. Recognizing the limitations of traditional connectivity and diagnosability, we investigate the connectivity and diagnosability of DPCell under the condition that each fault-free node in the network has at least $g$ fault-free neighbors. The results indicate that, under this condition, the connectivity and diagnosability of DPCell exceed its traditional metrics by more than $g$ times.