Minimizing AoI With Throughput Requirements in Multi-Path Network Communication

We consider a single-unicast networking scenario where a sender periodically sends a batch of data to a receiver over a multi-hop network, possibly using multiple paths. We study problems of minimizing peak/average Age-of-Information (AoI) subject to throughput requirements based on a stylized deterministic model in this scenario. The consideration of batch generation and multi-path communication differentiates our AoI study from existing ones. We first show that our AoI minimization problems are NP-hard, but only in the weak sense, as we develop an optimal algorithm with a pseudo-polynomial time complexity. We then prove that minimizing AoI and minimizing maximum delay are “roughly” equivalent, in the sense that any optimal solution of the latter is an approximate solution of the former with bounded optimality loss. We leverage this understanding to design a general approximation framework for our problems. It can build upon any $\alpha $ -approximation algorithm of the maximum delay minimization problem to construct an $(\alpha +\mathsf {c})$ -approximate solution for minimizing AoI. Here $\mathsf {c}$ is a constant depending on the throughput requirements. Furthermore, we show that our results can be extended to the multiple-unicast setting. Simulations over various network topologies validate the effectiveness of our approach. Our results make a major advance to optimizing AoI in multi-path communication, and hence can be of broad interest to the networking research community.

I. Introduction

Age-of-Information (AoI) is a critical networking performance metric for periodic services that require timely transmissions. Kaul et al. [3] measures the AoI as the time that elapsed since the last received update was generated. In this paper, we study AoI-minimization problems of supporting a periodic transmission task under a single-unicast networking scenario. Specifically, the task requires a sender to send a batch of data (packets) periodically to a receiver over a multi-hop network, possibly using multiple paths. We use a stylized deterministic model to model the network. Our objective is to minimize peak/average AoI subject to both a minimum and a maximum throughput requirement, by jointly optimizing throughput and multi-path routing strategy. We assume the amount of data in the batch is fixed, hence the throughput (the ratio of the volume of the data batch over the task activation period) only varies with the task activation period.