DecTE: Achieving Optimal Network Group Efficiency in Selfish Multidomain Networks

Newly emerging technological applications have resulted in a growing demand for high-quality network services. This demand has motivated recent research on employing centralized traffic engineering (TE) to improve the quality of service and network efficiency. However, the inherent distrust problem and information barrier make centralized TE infeasible in networks with selfish autonomous domains. In addition, the existing decentralized TE methods for multidomain scenarios have several drawbacks, which include nonoptimal network efficiency and the violation of domain selfishness in some integrated designs. Considering that selfish domains only maximize their domain efficiency (DE), we propose a decentralized TE method called DecTE, which achieves optimal network group efficiency (NGE) without sacrificing the DE of individual domains. We first establish a decentralized TE model with a revenue distribution mechanism. Through theoretical derivation, we prove the existence of an optimal revenue distribution (ORD) that makes selfish domains reach their Nash equilibrium and attain optimal NGE. Then, we propose decentralized algorithms to make the revenue distribution iteratively converge to the ORD, and per-domain online TE is performed based on the revenue distribution result. Simulation results demonstrate that DecTE outperforms the existing methods with a 10
$\%$
to 30
$\%$
NGE improvement under heavy traffic loads.