Overlays with Preferences: Distributed, Adaptive Approximation Algorithms for Matching with Preference Lists
Artikel i vetenskaplig tidskrift, 2013

A key property of overlay networks is the overlay nodes’ ability to establish connections (or be matched) to other nodes by preference, based on some suitability metric related to, e.g., the node’s distance, interests, recommendations, transaction history or available resources. When there are no preference cycles among the nodes, a stable matching exists in which nodes have maximized individual satisfaction, due to their choices, however no such guarantees are currently being given in the generic case. In this work, we employ the notion of node satisfaction to suggest a novel modeling for matching problems, suitable for overlay networks. We start by presenting a simple, yet powerful, distributed algorithm that solves the many-to-many matching problem with preferences. It achieves that by using local information and aggregate satisfaction as an optimization metric, while providing a guaranteed convergence and approximation ratio. Subsequently, we show how to extend the algorithm in order to support and adapt to changes in the nodes’ connectivity and preferences. In addition, we provide a detailed experimental study that focuses on the levels of achieved satisfaction, as well as convergence and reconvergence speed.

distributed

matching

overlay networks

adaptive

Författare

Georgios Georgiadis

Chalmers, Data- och informationsteknik, Nätverk och system

Marina Papatriantafilou

Chalmers, Data- och informationsteknik, Nätverk och system

Algorithms

19994893 (eISSN)

Vol. 6 4 824-856

Styrkeområden

Informations- och kommunikationsteknik

Transport

Energi

Ämneskategorier

Data- och informationsvetenskap

Elektroteknik och elektronik

Datavetenskap (datalogi)

DOI

10.3390/a6040824

Mer information

Skapat

2017-10-07