Internet Tomography in support of Internet and network simulation and emulation modelling

Abstract

Internet performance measurement data extracted through Internet Tomography techniques and metrics and how it may be used to enhance the capacity of network simulation and emulation modelling is addressed in this paper. The advantages of network simulation and emulation as a means to aid design and develop the component networks, which make up the Internet and are fundamental to its ongoing evolution, are highlighted. The Internet's rapid growth has spurred development of new protocols and algorithms to meet changing operational requirements such as security, multicast delivery, mobile networking, policy management, and quality of service (QoS) support. Both the development and evaluation of these operational tools requires the answering of many design and operational questions. Creating the technical support required by network engineers and managers in their efforts to seek answers to these questions is in itself a major challenge. Within the Internet the number and range of services supported continues to grow exponentially, from legacy and client/server applications to VoIP, multimedia streaming services and interactive multimedia services. Services have their own distinctive requirements and idiosyncrasies. They respond differently to bandwidth limitations, latency and jitter problems. They generate different types of "conversations" between end-user terminals, back-end resources and middle-tier servers. To add to the complexity, each new or enhanced service introduced onto the network contends for available bandwidth with every other service. In an effort to ensure networking products and resources being designed and developed handling diverse conditions encountered in real Internet environments, network simulation and emulation modelling is a valuable tool, and becoming a critical element, in networking product and application design and development. The better these laboratory tools reflect real-world environment and conditions the more helpful to designers they will be. This paper describes how measurement of empirical Internet data, obtained through Internet Tomography Measurement Systems (ITMS), can serve an important role in providing these laboratory simulation and emulation modelling tools with Internet parameterization data. The data being extracted from up-to-date real-world Internet can be used to recreate such conditions within the modelling experiments. This paper sets out how such data may be captured over extended and targeted periods of time and used in the laboratory modelling and experiments to define best-, average-, and worst-case Internet scenarios likely to be encountered by the applications or network upgrades being designed. An example of real-time one-to-many global-based Internet QoS measurement data sets obtained within a collaboration in the Réseaux IP Européens (RIPE) project for this purpose is presented.