A TCP Congestion Control Optimization Method for SDN-Based Data Center Networks
Jinlin Xu, Wansu Pan, Haibo Tan, Longle Cheng · 2024
Data centers can provide high aggregate bandwidth for large-scale network services, but traditional TCP protocols do not meet the requirements of data center environments. The emergence of software-defined networking (SDN) provides a global view of the data center network and enables centralized traffic control, which can effectively solve the data center congestion challenge. Taking advantage of the SDN architecture, we propose an adaptive congestion control method called E-DCTCP. It combines Explicit Congestion Notification (ECN) and Round- Trip Time (RTT) to model congestion awareness and ECN marking. It selects appropriate ECN markers based on queue lengths and their growth slopes and adjusts the Congestion Window (CWND) by calculating the RTT, which can alleviate false congestion caused by sudden flows. Evaluations performed with Mininet simulations show that E-DCTCP offers advantages in terms of throughput, flow completion time (FCT), latency, and prevention of packet loss. These benefits help reduce data center congestion and improve data transfer performance.