Dual Layer Scale Free Network Topology Synthesis
AuthorAkgun, Mehmet Burak
AdvisorGunes, Mehmet H
Computer Science and Engineering
StatisticsView Usage Statistics
Comprehensive analysis that aim to better understand the topology of real world networks and development of algorithms that replicate their characteristics have been an important research issue. Internet topology generators focus on generating synthetic graphs that attempt to imitate certain topological characteristics of the Internet at various levels. Although the accuracy of newly developed network protocols or algorithms do not depend on the underlying topology, the performance generally depends on the topology. As a result, network researchers have concentrated on generating representative synthetic topologies and widely utilize them to analyze the performance of their design in simulation or emulation environments. Topology generators typically represent the Internet topology as a graph composed of point-to-point links. In this dissertation, we focus on the multi-access links, i.e., link-layer networks that form the same collision domain between attached devices, as they should be considered to reflect the single hop connectivity of the devices attached to a subnetwork. While consideration of the multi-access links helps to better capture the underlying networks, this also introduces challenges in the generation of hypergraphs that accurately represent the Internet. In this dissertation, we discuss the implications of multi-access links on the synthetic network generation. Additionally, we analyze the large-scale characteristics of sampled Internet topology data sets and observe that in addition to the commonly analyzed node degree distribution, the subnet size and the router interface distributions exhibit power-law characteristics. Based on findings from the measurement studies, we introduce Subnet Oriented Network Topology generation that incorporates the observed measures to produce 2-mode Internet topologies. In particular, generated topologies capture the 2-mode relation between layer-2 (i.e., subnet and interface distributions) and layer-3(i.e., observed degree) that is missing from the current network generators that produce 1-mode graphs.