A Low-overhead and Real-time Service Recovery Mechanism for Data Center Networks with SDN
Ziqing Zhang, Jie Ma, Shuoyu Mao, Peng Liang, Qi Liu · 2024
In Data Center Software-Defined Networking (DCSDN), the expansion of network scale and the increase of network applications exacerbate network failures, which bring challenges to service continuity. Existing service recovery mechanisms include network fault detection strategies and fault mitigation strategies. However, the existing fault sensing strategies have low sensing accuracy and are difficult to meet the fine-grained requirements of fault detection. Sending sensing messages with a fixed periodicity will greatly consume link resources and increase the communication cost. Meanwhile, the fault mitigation strategy cannot flexibly respond to different service demands and real-time changing network states. To tackle the above issues, this paper proposes a real-time service recovery mechanism, which designs fault fields carrying fine-grained network state information on the programmable data plane, which enables monitoring of specific paths. Based on the detection of the network state, the control plane calculates the priority of the alternate path in advance. When a failure occurs, it can ensure that critical traffic is prioritized to be transmitted through the higher-priority alternate path, thus guaranteeing the continuity of network services. To verify the effectiveness of the proposed mechanism, we test it in a programmable network simulation environment, the results show that the mechanism can detect network faults with low overhead and switch alternate paths quickly and accurately when faults occur, providing an effective real-time recovery mechanism for the network.