Minimizing Age-of-Information with Throughput Constraints in Multi-Path Network Communication
share
Age-of-information (AoI) is a newly proposed time-critical networking performance metric. In this paper, we consider multi-path network communication problems for fulfilling a single-unicast periodic transmission task. The task generates certain amount of data at a sender periodically, and streams them to a receiver using multiple paths over a multi-hop network. Our objective is to minimize either the peak AoI or the average AoI, by jointly optimizing the multi-path routing strategy and the task activation length. To our best knowledge, (i) as compared to existing AoI studies, we are the first to minimize the task-level AoI with the multi-path routing optimization involved, and (ii) as compared to existing time-critical multi-path communication studies, we minimize the AoI with the task activation length to be a variable, while they minimize the maximum delay given a fixed task activation length. In this paper, first, comparing the newly proposed AoI with the common-studied maximum delay, we show that the maximum-delay-optimal solution can achieve suboptimal AoI, and we characterize near-tight optimality gaps of this suboptimal AoI. Second, we give fundamental structures of our problems, by proving that the peak-AoI minimization can differ from the average-AoI minimization, but both of them are weakly NP-hard, with a pseudo-polynomial time exact algorithm designed. Third, we develop an approximation framework, which can leverage an arbitrary approximation algorithm of the maximum-delay-minimization problem to solve our AoI-minimization problems approximately. Our framework has a constant approximation ratio, and a time complexity that is the same as that of the used approximation algorithm. Finally, we conduct extensive random simulations to empirically verify our proposed results and evaluate our designed algorithms.
READ FULL TEXT