Abstract

Applications like inter data-centre synchronisation or client-to-cloud backups require a reliable end-to-end data transfer, however, they typically do not have strong capacity or latency constraints, just a loose delivery deadline. Besides, their potential to disrupt more quality-constrained flows should be kept to a minimum. These applications could be well served by a transport protocol providing a less-than-best-effort (LBE) or scavenger service rather than TCP but, neither TCP nor standard LBE methods like LEDBAT consider any notion of deadline or completion time. TCP simply tries to maximise the use of available capacity, while LEDBAT tries to enforce an LBE behaviour regardless of any timeliness requirements. This paper introduces a framework for adding both LBE behaviour and awareness of “soft” delivery deadlines to any congestion control (CC) algorithm, whether loss-based, delay-based or explicit signaling-based. This effectively allows it to turn an arbitrary CC protocol into a scavenger protocol that dynamically adapts its sending rate to network conditions and remaining time before the deadline, to balance timeliness and transmission aggressiveness. Network utility maximization (NUM) theory provides a solid foundation for the proposal. The effectiveness of the approach is validated by numerical and simulation experiments, with TCP Cubic and Vegas used as examples.