Weighted Shortest Path Models: A Revisit to the Simulation of Internet Routing

TL;DR

This paper introduces a novel weighted shortest path model for simulating Internet routing, which better captures real-world routing behaviors by assigning link weights based on local information, improving upon traditional unweighted models.

Contribution

It proposes a new local information-based weighted shortest path model for Internet routing simulation, capturing routing behaviors more accurately than existing unweighted models.

Findings

The model accurately replicates Internet route length distribution.

Optimal parameters suggest packets prefer nodes with smaller degrees.

Routing behavior balances efficiency and congestion.

Abstract

Understanding how packets are routed in Internet is significantly important to Internet measurement and modeling. The conventional solution for route simulation is based on the assumption of unweighted shortest path. However, it has been found and widely accepted that a packet in Internet is usually not transmitted along the unweighted shortest path between its source and destination. To better simulate the routing behavior of a packet, we thoroughly explore the real-world Internet routes and present a novel local information based simulation model, with a tuning parameter, which assigns weights to links based on local information and then simulates the Internet route with weighted shortest path. Comparisons with baseline approaches show its capability in well replicating the route length distribution and other structural properties of the Internet topology. Meanwhile, the optimal parameter of this model locates in the range of (0, 2), which implies that a packet inside the Internet inclines to move to nodes with small degrees. This behavior actually reflects the design philosophy of Internet routing, balancing between network efficiency and traffic congestion.