Network Path Capacity Comparison without Accurate Packet Time Information

A fundamental problem of current bandwidth estimation methods is that they require accurate packet time information. However, it is hard to accurately measure packet time information in an increasing number of network environments, such as widely deployed high speed networks, and emerging cloud computing networks. Motivated by the observation that many applications only need the relative bandwidth information of different paths instead of the actual bandwidth information of a single path, we propose sequence-based bandwidth comparison. Specifically, this paper proposes a capacity comparison method, called Path Comp, which can relatively compare the capacities of the paths from two senders to the same receiver. Path Comp mainly uses the arrival sequence information of packets, and does not require any accurate packet time information. Our test bed, campus network, and EC2 experiments show that Path Comp can not only determine which path is faster but also accurately determine how much faster in a variety of network environments.