As many-to-one traffic patterns prevail in data center networks, TCP flows often suffer from severe unfairness in sharing bottleneck bandwidth, which is known as the TCP outcast problem. The cause of the TCP outcast problem is the bursty packet losses by a drop-tail queue that triggers TCP timeouts and leads to decreasing the congestion window. This paper proposes TCPRand, a transport layer solution to TCP outcast. The main idea of TCPRand is the randomization of TCP payload size, which breaks synchronized packet arrivals between flows from different input ports. We investigate how TCPRand reduces consecutive packet drops and demonstrate various benefits of TCPRand with extensive experiments and ns-3 simulation. Our evaluation results show that TCPRand guarantees the superior enhancement of TCP fairness with negligible overheads in all of our test cases.

In this paper, we suggest a efficient multiple-forwarder based file distribution scheme working over a tree-based overlay multicast. The proposed scheme takes advantage of a fact that each member joins the session with time interval. In a tree, the set of data delivered from a parent to each member since a member has joined until the next member joins is not overlapped. By using this property, in the proposed scheme, each forwarder distributes data to its succeeding members regardless of the underlying tree. The effectiveness of the proposed scheme is measured using numerical analysis and the performance evaluation on planet-lab shows that the proposed scheme enhances the throughput while maintaining data duplication ratio and control overhead significantly less compared to an existing scheme, Bullet.

There have been many research efforts to enhance the throughput of file distribution using overlay multicast. The tree-based approach has limitations in enhancing throughput due to its single forwarder architecture. Recent approaches introduce multiple packet forwarders other than the parent in the delivery tree for throughput enhancement. However, the overhead of control packet exchanges is required to appoint proper forwarders and to avoid duplicate packet delivery. In this paper, we suggest a throughput enhancement scheme for a tree-based overlay multicast scheme with multiple forwarders. It exploits a forward error correction scheme and the fact that the members join the session with time interval. Thus, each set of packets that are delivered from a parent to a member after the member joins and before a succeeding member joins is not overlapped. By deploying a dedicated sender to manage the membership information and packet reception state, the overhead of control packet exchanges is alleviated significantly. The performance evaluation shows that the proposed scheme reduces file delivery time with low control overheads and small packet duplication ratio, compared with existing scheme, Bullet.