Citation Context

...ation and evaluation of network topologies. In the current environment, degree distributions and other large-scale statistics are popular metrics for evaluating how representative a given topology is =-=[42]-=-, and proposed topology generators are often evaluated on the basis of whether or not they can reproduce the same types of macroscopic statistics, especially power law-type degree distributions [11]. ...

Citation Context

...1: ISP topologies as measured by Rocketfuel. Random topologies: In addition to real topologies, for diversity we also randomly generate power-law topologies using BRITE [25], since a number of papers =-=[11, 39]-=- have shown that the power-laws capture the Internet structure quite well. We generate 100-node router-level topologies with edge density (i.e., the number of neighboring nodes that each new node conn...

Citation Context

...D WORK Our work draws on and benefits from three separate lines of work – topology collection and analysis, policy inference, and studies of the impact of routing policies. Various researchers have collected and analyzed Internet topology at different granularities, such as router-level [15, 7, 27] and AS-level [9, 6]. We use the techniques of router-level topology collection (traceroute) and router name-to-location mapping [27, 22] to collect POP-level topologies of 65 major ISPs and their interconnections. Extending previous work that studied topology properties, such as degree distribution [9, 35, 5] and resilience [35], we study how topology impacts path inflation. Inference of routing policy to date has been limited to the interdomain case [12, 29]. We add two key missing pieces to routing policy inference work: inference of intra-domain traffic engineering and inference of peering policy between a pair of ISPs. Starting with the Detour project [26], much attention has been given to the impact of inter-domain routing policies on performance. Detour showed that there exist paths in the Internet that are better than those being used. While it measured the end-to-end impact of routing poli...

Citation Context

...d to the point where realistic router-level maps are now becoming available [3, 6, 11, 19]. These maps are in turn being used to study the structure of the Internet (e.g., node outdegree distribution =-=[7, 21]-=-). However, we note that the current maps do not support studies based on Internet paths. This is because topology alone does not determine which paths will be selected. Rather, the routing model, alo...

Citation Context

... be used for trace-driven simulation of their proposed search mechanisms. Finally, the research studies on characterization of the Internet topology (e.g., [8]) and network topology generators (e.g., =-=[34]-=-) are closely related to our work. However, these studies focus on the Internet topology rather than an overlay topology. We plan to conduct further characterization of the Gnutella topology by applyi...

Citation Context

...the relationship between à and è can be well captured by àésèzffêèÝë , where ì ß . is the dimension of the grid network. In a tree or a random graph, àísèzff«îì&ï best describes their relationship =-=[22]-=-. Here ì ß . is related to the connectivity of the random graph. DIHA - Distributed Indexing with Hashing Architecture In this architecture, each peer is responsible for a subset of the index space....