AEGIS: Throughput-Guaranteed Resilient Routing via a Conditional Value-at-Risk Approach

Thu, 16 Apr 2026 00:00:00 +0000

Abstract

The past decade has witnessed significant progress in next-generation wireless networks. Resilient routing is essential for maintaining reliability in mission-critical network services, particularly in dynamic and adversarial environments. Traditional traffic engineering (TE) approaches rely on pre-computed paths. Still, they face performance limitations when the number of pre-computed paths is small and scalability challenges when the number is large. This study seeks to answer the fundamental question: “How can we achieve throughput-guaranteed resilient routing under network failures without pre-computing routing paths?” We propose AEGIS, a novel throughput-guaranteed resilient routing scheme leveraging a conditional value-at-risk (CVaR) approach, which proactively guarantees the required throughput under normal conditions and enables recovery during network failures. Specifically, we propose an optimization problem that minimizes the CVaR of total throughput loss across all the failure situations while respecting user budget and network constraints. The above optimization problem is non-differentiable and non-linear; we then reformulate it as an equivalent linear program (LP) and develop an optimal solution. However, the above solution will induce cyclic flows due to resource reservation behaviors. To achieve a more resource-efficient routing, we propose a bisection approach to obtain a CVaR upper bound so that the corresponding routing is acyclic. Extensive numerical evaluations demonstrate the trade-offs among various approaches and highlight the advantages of AEGIS.

Dynamic Flow Routing and Scheduling for Time-Critical Network Services

Mon, 06 Oct 2025 00:00:00 +0000

Abstract

A key challenge in next-generation networks is providing intelligent network services to support time-sensitive applications such as AR/VR and the Internet of Military Things (IoMT). This work aims to answer the question: “How can we design a network service scheduling and flow routing scheme to satisfy the stringent demand and deadline requirements of tactical network applications?” We take a time-expanded graph-based approach to this study. Specifically, given a time-expanded graph constructed from an original network graph, we propose an optimization problem to find a subset of network services to maximize their total completion rewards such that the data of all services in this subset can be transmitted to their destination nodes by their deadlines. Unfortunately, solving this problem directly is challenging since it is a mixed-integer program. We propose a heuristic algorithm based on a sequential rounding approach where a linear programming (LP) feasibility problem is iteratively solved to update the network service subset, guided by a relaxed reward maximization problem that prioritizes high-reward requests. To further improve efficiency, we fine-tune the above LP feasibility problem from a static flow perspective, enabling fast feasibility checks on the original graph, and design the corresponding efficient heuristic algorithm. The simulation results demonstrate the trade-off among different approaches and validate the effectiveness of the proposed algorithms.

Network Optimization | Xuanli Lin

AEGIS: Throughput-Guaranteed Resilient Routing via a Conditional Value-at-Risk Approach

Abstract

Dynamic Flow Routing and Scheduling for Time-Critical Network Services

Abstract