Investigation of Means of Mitigating Congestion in Complex, Distributed Network Systems by Optimization Means and Information Theoretic Procedures
Abstract
This work investigates how the values of network metrics affect congestion in wireless ad-hoc networks. The metrics considered in this work are average path length, average degree, clustering coefficient, and offdiagonal complexity. Based on the levels of these metrics, insight is provided on the clustering algorithm to choose what will minimize congestion. Congestion is evaluated using a node betweenness measure and the candidate clustering algorithms are lowest ID, highest degree, and MOBIC. To obtain data for analysis, a network simulator was developed using Microsoft Visual C++ 2005. The simulator is capable of creating networks of varying complexity, clustering these networks using the aforementioned algorithms, and evaluating each of the five metrics. Analysis of the results confirmed that congestion levels increase with complexity. This was evidenced by evaluation of all five network metrics. Also, networks with relatively low levels of complexity will have minimal congestion, regardless of the clustering method used.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 2008
- Accession Number
- ADA483777
Entities
People
- Frank Mufalli
- Jim Llinas
- Rakesh Nagi
- Sumita Mishra
- W. F. Lawless
Organizations
- Calspan-University of Buffalo Research Center