How to build a sovereign network? -- A proposal to measure network sovereignty

TL;DR

The paper addresses network sovereignty, defined as operating without dependencies on a single manufacturer, and introduces the Cut Set Coloring (CSC) framework with the MCS-CSC, Flow-CSC, and Network-CSC scores. It then presents CSC-ILP, an ILP that assigns network nodes to manufacturers to maximize the minimum Flow-CSC across all flows under a cost constraint $C_T$ with per-manufacturer costs $C_m$. Through case studies on four sndlib topologies and comparisons with DAP and centrality-based heuristics, the authors demonstrate that CSC-ILP achieves superior worst-case and distributional sovereignty metrics, and they provide practical guidelines for operator deployment and manufacturer mapping. The work highlights CSC's scalability and its potential to guide multi-vendor network design, while noting limitations such as subcomponent sharing and probabilistic manufacturer failures for future integration with reliability metrics.

Abstract

Network sovereignty is a network operator's ability to reduce the dependency on component manufacturers to minimize the impact of manufacturer failures. Network operators now face new design challenges to increase network sovereignty and avoid vendor lock-in problems because a high dependency on a manufacturer corresponds to low survivability if that manufacturer is unavailable. The main contribution of this work is the proposal of a novel metric to measure network sovereignty, the Cut Set Coloring (CSC) score. Based on the CSC core metric CSC-ILP, our Integer Linear Program formulation is presented to maximize network sovereignty. We compare CSC-ILP's performance with state of the art manufacturer assignment strategies.

Figures (3)

• Figure 1: Network sovereignty vs. availability example, using a DCN topology and its RBD. $A_f$ and $A_n$ denote the flow and node availabilities, respectively. $A_{x||y}$ denotes the availability of two nodes $x$ and $y$ in parallel. Each node color represents a different manufacturer.
• Figure 2: Network-CSC score represented as Flow-CSC score distribution vs. Number of manufacturers ($|M|$) for Polska topology. The proposed CSC-ILP assignments outperform DAP, Naga, Min cut set heuristic, and centrality metrics. Flow-CSC score increases with more manufacturers. The star in the box plots represents the mean. The heights of different color bars indicate the distribution of the central 50% of the Flow-CSC scores. The lines on the top and the bottom of the boxes denote the remaining data range on the top and bottom quartiles, respectively. The white circles denote outliers. The absence of a box shows that the Flow-CSC scores are concentrated near the mean.
• Figure 3: CSC-ILP's performance: Polska with five manufacturers ($|M| = 5$), CSC-ILP outperforms other assignments.