SpeedCam: Towards Efficient Flow Monitoring for Multipath Communication

Network monitoring is essential for traffic engineering, maintenance, and troubleshooting purposes and thus forms an integral part of network management. However, observing each and every packet may not be feasible or at least be very costly. It is therefore crucial for network operators to ensure a scalable and efficient monitoring. With the emergence of multipath communication as facilitated by new network architectures like SCION, monitoring becomes an even more challenging task. In a multipath network, operators need to be able to monitor their customers' traffic flows across different network paths to ensure optimal network utilization, fault tolerance and fairness. Traditional single-path flow monitoring such as sampling-based mechanisms fall short, since packets may be spread across a potentially large number of different paths. To address this challenge, this paper proposes SPEEDCAM, a new approach that aims to achieve scalable and efficient flow monitoring in multi-path networks. Our approach which is based on probabilistic probe selection significantly reduces the number of required monitoring probes, while enabling an effective flow information gathering. With an implementation of SPEEDCAM in the SCIONLab network, we demonstrate more than 89% of monitoring accuracy with a small fraction of network routers covering only 50% of network traffic in the multipath network.