[Paper Review] A Graph Theoretical Approach to Network Encoding Complexity

Li Xu, Weiping Shang+1 more|arXiv (Cornell University)|Feb 3, 2012

Interconnection Networks and Systems8 references4 citations

TL;DR

This paper introduces a graph-theoretic framework to analyze network encoding complexity by modeling mergings of Menger's paths in acyclic directed networks. By defining constants $\mathcal{M}(c_1,\dots,c_l)$ and $\mathcal{M}^*(c_1,\dots,c_l)$ that bound the minimum number of path mergings across multiple sessions, the authors derive tight bounds and exact values for encoding complexity in multicast, two-way, and unicast networks using path concatenation and structural decomposition techniques.

ABSTRACT

Consider an acyclic directed network $G$ with sources $S_1, S_2,..., S_l$ and distinct sinks $R_1, R_2,..., R_l$. For $i=1, 2,..., l$, let $c_i$ denote the min-cut between $S_i$ and $R_i$. Then, by Menger's theorem, there exists a group of $c_i$ edge-disjoint paths from $S_i$ to $R_i$, which will be referred to as a group of Menger's paths from $S_i$ to $R_i$ in this paper. Although within the same group they are edge-disjoint, the Menger's paths from different groups may have to merge with each other. It is known that by choosing Menger's paths appropriately, the number of mergings among different groups of Menger's paths is always bounded by a constant, which is independent of the size and the topology of $G$. The tightest such constant for the all the above-mentioned networks is denoted by $\mathcal{M}(c_1, c_2,..., c_2)$ when all $S_i$'s are distinct, and by $\mathcal{M}^*(c_1, c_2,..., c_2)$ when all $S_i$'s are in fact identical. It turns out that $\mathcal{M}$ and $\mathcal{M}^*$ are closely related to the network encoding complexity for a variety of networks, such as multicast networks, two-way networks and networks with multiple sessions of unicast. Using this connection, we compute in this paper some exact values and bounds in network encoding complexity using a graph theoretical approach.

Motivation & Objective

• To establish a graph-theoretic model for analyzing network encoding complexity based on edge-disjoint Menger paths and their mergings.
• To define and compute the minimal number of mergings among Menger paths from different source-sink pairs, denoted by $\mathcal{M}$ and $\mathcal{M}^*$, which directly relate to encoding complexity.
• To derive tight lower bounds on encoding complexity for various network types, including multicast, two-way, and multi-session unicast, using structural concatenation of subnetworks.
• To formalize the connection between path mergings and the necessity of network coding, showing that mergings act as bottlenecks requiring coding to avoid congestion.

Proposed method

• Model acyclic directed networks with sources $S_i$, sinks $R_i$, and define $c_i$ as the min-cut between $S_i$ and $R_i$.
• Define a merging as a shared subpath where at least two Menger paths from different source-sink pairs converge at an edge and continue together.
• Use Menger’s theorem to ensure existence of $c_i$ edge-disjoint paths from $S_i$ to $R_i$, forming a group of Menger’s paths.
• Introduce $|G|_{\mathcal{M}}(\alpha_1,\dots,\alpha_l)$ as the count of distinct merging edges across all path groups, without multiplicity.
• Construct non-reroutable graphs via concatenation: reverse edges in subgraphs, split sources and sinks, and identify nodes to form larger networks.
• Apply recursive concatenation of $k-1$ subgraphs to build an $(n_1,\dots,n_k)$-graph, preserving merging counts and enabling inductive bounds.

Experimental results

Research questions

• RQ1 What is the minimal number of mergings among Menger paths from different source-sink pairs in a network with given min-cuts?
• RQ2 How do the constants $\mathcal{M}(c_1,\dots,c_l)$ and $\mathcal{M}^*(c_1,\dots,c_l)$ relate to network encoding complexity in multicast and two-way communication?
• RQ3 Can the number of mergings be bounded independently of network size and topology, and what are the tightest such bounds?
• RQ4 How can path concatenation techniques be used to derive lower bounds on encoding complexity for multi-session networks?
• RQ5 What structural properties of networks force network coding, and how are these related to merging points of Menger paths?

Key findings

• The paper establishes that the number of mergings among Menger paths from different source-sink pairs is bounded by a constant independent of network size and topology.
• For networks with $l$ distinct sources and sinks, the minimal number of mergings is denoted $\mathcal{M}(c_1,\dots,c_l)$, and for identical sources, it is $\mathcal{M}^*(c_1,\dots,c_l)$.
• The paper proves $\mathcal{M}(n,n) \geq \mathcal{M}^*(n+1,n+1) + \mathcal{M}^*(n-1,n-1) + (n-1)$, providing a recursive lower bound for encoding complexity.
• For $n_1 \leq \cdots \leq n_k$, the inequality $\mathcal{M}^*(n_1,\dots,n_k) \geq \sum_{i=1}^{k-1} \mathcal{M}^*(n_i,n_i)$ is derived, showing that complexity grows with the number of sessions.
• The authors demonstrate that mergings act as bottlenecks requiring network coding, and the number of such points directly determines the minimum encoding cost.
• Through graph concatenation, the paper constructs non-reroutable networks with controlled merging counts, enabling exact computation and tight bounds on encoding complexity.