A framework for delay emulation of large-scale internetworks

Abstract

Several network applications require access to a large number of geographically distributed machines for testing and debugging purposes. Typically, it is the case that either access is not available for so many machines, or even if access were available, it may be infeasible to perform tests in such a complex setting. This thesis proposes a framework - Scalable Internetwork Delay Emulation Framework (SIDEF) - that allows modeling of such complex network scenarios in a simple laboratory environment. The framework models and dynamically adapts the test traffic according to the network delay characteristics observed on the Internet in real-time. The proposed framework consists of three main modules: Virtual Host Configuration Module, Delay Estimator, and Delay Injector. The Virtual Host Configuration Module enables a machine to act as a set of virtual hosts with arbitrary IP addresses. The Delay Estimator can compute delay values for any two arbitrary IP addresses. The Delay Injector captures packets at the IP layer and delays them according to the estimates provided by the Delay Estimator. We developed a prototype implementation - Scalable Internetwork Delay Emulator (SIDE) - which is based on this architecture. Experimental results show that the proposed framework is effective for testing large-scale network applications.