Distributed Strong Diameter Network Decomposition

TL;DR

This paper presents a novel distributed algorithm that constructs strong network decompositions with logarithmic parameters in polylogarithmic time, solving a longstanding open problem in distributed graph algorithms.

Contribution

It introduces a method to compute strong (O(log n), O(log n)) network decompositions in O(log^2 n) time, improving upon previous bounds and establishing new capabilities for distributed graph partitioning.

Findings

Strong (O(log n), O(log n)) network decompositions can be computed in O(log^2 n) time.

The paper provides a tradeoff between the parameters of the network decomposition.

The approach adapts the shifted shortest path method for distributed computation.

Abstract

For a pair of positive parameters $D,\chi$, a partition ${\cal P}$ of the vertex set $V$ of an $n$-vertex graph $G = (V,E)$ into disjoint clusters of diameter at most $D$ each is called a $(D,\chi)$ network decomposition, if the supergraph ${\cal G}({\cal P})$, obtained by contracting each of the clusters of ${\cal P}$, can be properly $\chi$-colored. The decomposition ${\cal P}$ is said to be strong (resp., weak) if each of the clusters has strong (resp., weak) diameter at most $D$, i.e., if for every cluster $C \in {\cal P}$ and every two vertices $u,v \in C$, the distance between them in the induced graph $G(C)$ of $C$ (resp., in $G$) is at most $D$. Network decomposition is a powerful construct, very useful in distributed computing and beyond. It was shown by Awerbuch \etal \cite{AGLP89} and Panconesi and Srinivasan \cite{PS92}, that strong $(2^{O(\sqrt{\log n})},2^{O(\sqrt{\log n})})$ network decompositions can be computed in $2^{O(\sqrt{\log n})}$ distributed time. Linial and Saks \cite{LS93} devised an ingenious randomized algorithm that constructs {\em weak} $(O(\log n),O(\log n))$ network decompositions in $O(\log^2 n)$ time. It was however open till now if {\em strong} network decompositions with both parameters $2^{o(\sqrt{\log n})}$ can be constructed in distributed $2^{o(\sqrt{\log n})}$ time. In this paper we answer this long-standing open question in the affirmative, and show that strong $(O(\log n),O(\log n))$ network decompositions can be computed in $O(\log^2 n)$ time. We also present a tradeoff between parameters of our network decomposition. Our work is inspired by and relies on the "shifted shortest path approach", due to Blelloch \etal \cite{BGKMPT11}, and Miller \etal \cite{MPX13}. These authors developed this approach for PRAM algorithms for padded partitions. We adapt their approach to network decompositions in the distributed model of computation.