Improving ECN marking scheme with micro-burst traffic in data center networks

In data centers, micro-burst is a common traffic pattern. The packet dropping caused by it usually leads to serious performance degradations. Therefore, much attention has been paid to avoiding buffer overflow caused by micro-burst traffic. In particular, ECN is widely used in data centers to keep persistent queue occupancy low, so that enough buffer space can be available as headroom to absorb micro-burst traffic. However, we find that instantaneous-queue-length-based ECN may cause problems in another direction - buffer underflow. Specifically, current ECN marking scheme in data centers is easy to trigger spurious congestion signals, which may result in overreaction of senders and queue length oscillations in switches. Since ECN threshold is low, the buffer may underflow and link capacity is not fully used. In this paper, we reveal this problem by experiments and simulations. Besides, we theoretically deduce the amplitude of queue length oscillations. The analysis result shows that overreaction of senders is caused by ECN mis-marking. Therefore, we propose Combined Enqueue and Dequeue Marking (CEDM), which can mark packets more accurately. Through simulations, we show that CEDM can greatly reduce throughput loss and improve flow completion time.