ICTON 2016 We.B3.2 978-1-5090-1466-8/16/$31.00 ©2016 IEEE 1 Traffic Generation for Telecom Cloud-Based Simulation Alba P. Vela * , Anna Vía, Fernando Morales, Marc Ruiz, and Luis Velasco Universitat Politècnica de Catalunya (UPC), Barcelona, Spain *e-mail:apvela@ac.upc.edu ABSTRACT With the incremental amount of applications running over the telecom cloud architecture it is becoming of paramount importance being able to run simulations aiming at evaluating the performance of such applications. To that end, one of the key elements in the simulation is how to generate network traffic. In this paper we propose realistic traffic functions that can be used for such purposes and present how those functions have been integrated in our OMNET++-based simulator. Keywords: traffic generation, network simulation, traffic profiles. 1. INTRODUCTION Traffic generation is a useful technique that enables studying and evaluating the network performance through simulation, when real traffic traces are not available. In fact, as pointed out by authors in [1], there is an absence of public availability of real world network traces, specifically for traffic in the operators’ and Internet service providers core transport networks. Therefore, due this unavailability, one option is to attempt generating simulated traces which resemble to real ones. Authors in [2] state that it is quite easy to generate traffic, but it is far more difficult to produce traffic that exhibit real characteristics, such as the ones observed through the Internet. Several traffic generators have been developed until present but, to the best of our knowledge, literature is mainly focused on generating representative IP traffic for packet-based networks or connection arrival based on the Poisson distribution for circuit-switched networks [6]. However, to evaluate the performance of operators’ network architectures, an intermediate traffic generation is needed in between packet generation and connection arrival modelling to reproduce continuous traffic. In this paper we face this problem and provide details of several traffic generation models to reproduce common traffic profiles. In addition, we present a framework that we integrate into the OMNeT++ discrete event-driven simulator, where its main feature is its modularity. 2. TRAFFIC PROFILES GENERATION This section presents general traffic profiles and the way they can be generated. Let us denote f(t;T) as the periodic function with period T returning the mean value of the model complex against time traffic. This function f(t;T) can be generated in multiple ways. For instance, one can define a piecewise linear function, i.e., a function composed of a number of linear segments, aiming to reproduce the traffic behavior against time. Another way could consist on a polynomial of some degree, or even a summation of functions, e.g., trigonometric sines. In addition, some random values around the average value are usually observed as a result, among others, of the monitoring process. That random function ε t can be modeled as a probability distribution function, e.g., following the normal (Gaussian) distribution. In such case, ε t ~ (µ, σ 2 ) defined by the mean µ and the standard deviation σ. In consequence, the complex traffic profile Y(t;T) to be reproduced can be generated as: (; ) (; ) t YtT ftT ε = + (1) Let us now present three different traffic profiles, where f(t;T) functions were generated using piecewise linear functions and ε t ~ (0,σ 2 ) (Fig. 1). Time (hh) Normalized traffic 0 0.2 0.4 0.6 0.8 1 0 2 4 6 8 10 12 14 16 18 20 22 0 0 2 4 6 8 10 12 14 16 18 20 22 0 Time (hh) Time (hh) 0 2 4 6 8 10 12 14 16 18 20 22 0 Figure 1. Traffic profiles generated: (a) Business, (b) CDN, and (c) DC2DC.