Table 2 Input parameters for default model
Parameter, symbol, description | Value | Explanation |
|---|---|---|
Number of ASes N | 500 | Limited by simulation time constraints |
Number of geographic locations G M a x | 50 | Based on approximate ratio of IXPs to peering networks in the Internet. PeeringDB ratio 10.27. GENESIS ratio 10.0 (PeeringDB 2012) |
Geographic expanse distribution | Zipf(1.6) | Based on data about number of participants at each IXP collected from PeeringDB (PeeringDB 2012). G(x) assigned randomly to each node |
Maximum expanse for an AS | 15 | |
Generated traffic distribution | Zipf(1.2) | Produces a heavy-tailed distribution of outgoing traffic. With this distribution, 0.1% of the ASes generate nearly 28% of the total traffic. This is consistent with the behavior reported in (Chang et al. 2005; Feldmann et al. 2004) & Labovitz et al. (2010), which show that the traffic produced by high-ranking ISPs and content providers follows a Zipf distribution. V G (x) assigned randomly to each node |
Consumed traffic distribution | Zipf(0.8) | Produces heavy-tailed distribution of incoming traffic, similar to measured traffic distribution at Georgia Tech. |
Mean consumed traffic | 500 Mbps | V C (x)∝|G(x)|, rationale being that a node with large expanse will also have a large number of access customers |
Private peering threshold Ω | 50 Mbps | Survey of peering strategies (Peering Strategy Survey 2011) |
Transit cost multiplier range P(x) | <DOLLAR/>[35,45]/Mbps per iteration | Parameterized based on IP transit prices recorded by Telegeography (Telegeography 2012). P(x) assigned randomly |
Transit cost exponent τ | 0.75 | Parameterized based on data from Chang et al. (2006) and Peering Strategy Survey (2011) |
Peering cost multiplier α | <DOLLAR/>20/Mbps per iteration | |
Peering cost exponent β | 0.40 | |
Selective peering ratio σ | 2.0 |