Minimizing the Maximum End-to-End Network Delay: Hardness, Algorithm, and Performance

Qingyu Liu,Lei Deng,Haibo Zeng,Minghua Chen
2017-01-01
Abstract:We consider the scenario where a source streams a flow at a fixed rate to a receiver across the network, possibly using multiple paths. Each link has a finite capacity. Transmission over a link incurs an integer delay if the rate is within the link capacity, and an unbounded delay otherwise. The objective is to minimize the maximum end-to-end delay experienced by the flow. The problem, denoted as Min-Max-Delay, appears in various practical scenarios, e.g., delay-critical video conferencing using inter-datacenter networks. In this paper, we first show Min-Max-Delay is weakly NP-hard and develop an exact algorithm with pseudo-polynomial time complexity. We then propose a Fully Polynomial Time Approximation Scheme (FPTAS). These results reveal fundamental difference between Min-Max-Delay problem and a similar maximum latency problem studied in the literature, for which is APX-hard and no PTAS exists unless P = NP. Moreover, there exists no exact pseudo-polynomial-time algorithm or constant-approximate algorithm for the maximum latency problem. We evaluate our algorithms in the scenario of routing delay-critical video-conferencing traffic over multiple paths of inter-datacenter networks, using simulations based on Amazon EC2 topology. Our algorithms both achieve the optimal maximum delay performance in all simulation instances, consistently outperforming all state-of-the-art solutions which only obtain sub-optimal maximum delay performance in certain instances. Furthermore, simulation results show that our achieved optimal delay always meet the delay requirement for video conferencing, while sub-optimal delay obtained by the alternatives fail to satisfy video-conferencing delay requirement for up to 15% of simulation instances between certain cross-continental source-receiver pair.
What problem does this paper attempt to address?