In pursuit of massive service emulation: a methodology for testbed building

Abstract

Every day, more services in the Internet (Youtube, Facebook, etc.) are being massively used, while network operators struggle to maintain their performance. Since simulations are not always accurate and evaluation prior to deployment is a must, it is often convenient to build an emulation testbed to test these services in more realistic environments. This methodology proposes four steps to guide researchers in this task. Before running the experiments and gathering results, it is necessary to specify the service under test, define the testbed requirements, and allocate the resources properly. For this reason, a set of techniques are proposed to solve the most common challenges at every step, such as adding real users or selecting the best simulation technique for our goals. The purpose is to efficiently employ the available resources without compromising the realism and accuracy of results. The application of the methodology to a case study, a high-quality audio and video on-demand service over a communication network, shows how resource constraints can be overcome to accomplish our goals.