Efficient and realistic generation of IP addresses

Abstract

Network simulation and emulation environments play a crucial role in evaluating proposed protocols, applications, and networked systems. In such settings, the ability to scalably and efficiently generate traffic that has characteristics similar to those measured in the live Internet is of great importance. A key aspect of generating realistic traffic is to assign source and destination IP addresses to traffic flows such that the statistical structure of the addresses is similar to what would be seen in a live Internet setting.In this paper, we propose and evaluate an algorithm and data structure for efficient and realistic generation of IP addresses. We describe our new method and compare it with existing and prior work, while also showing that our technique is far more efficient --- both in terms of memory consumed and computation time required. We also show that the statistical structure of the generated addresses is similar to what would be measured in the live Internet. Our results show that it is possible to efficiently generate addresses over the entire IPv4 address space, and that it is feasible to generate addresses from a/64 IPv6 subnet.