Randomizing TCP payload size for TCP fairness in data center networks

Abstract

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. Based on the current congestion window size and the CUBIC's congestion window growth function, TCPRand adaptively determines the proper level of randomness. With extensive ns-3 simulations and experiments, we show that TCPRand guarantees the superior enhancement of TCP fairness by reducing the TCP timeout period noticeably even in an environment where serious TCP outcast happens. TCPRand also minimizes the total goodput loss since its adaptive mechanism avoids unnecessary payload size randomization. Compared with DCTCP, TCPRand performs fairly well and only requires modification at the transport layer of the sender which makes its deployment relatively easier.