Modelado y analisis transitorio de un sistema de distribucion de video bajo demanda

Resumen

To achieve the provision of end-to-end Quality of Service (QoS) for multimedia services, the management of Service Level Agreements (SLAs) is required. Nevertheless, additional mechanisms are needed to quantitatively evaluate the user-level SLA parameters. The necessity to manage the QoS has special relevance in adaptive Video-on-Demand (VoD) systems. In these systems the bit rate is adjusted according to the available network resources. To achieve this, the QoS provided to the user is changed whenever is necessary. Both the service providers and the customers are indeed interested in tools capable of quantifying the performance of these systems from their respective points of view. This would allow them to establish the SLA with more easiness and transparency. In this research work of Doctoral Thesis, we have investigated analytical tools for the evaluation, a priori, of end-to-end QoS in adaptive VoD services. As a contribution to solve the lack of tools that could allow us to quantify performance metrics of these systems, the advantages of the so-called Performability techniques have been investigated. In particular, we have studied the Markov Reward Chains (MRC) and its evaluation using the randomization method. This investigation, along with the study of the adaptive VoD systems, is the base of the proposal of a generic methodology for the construction of Markovian models applicable to the adaptive multimedia services. The obtained models capture the dynamism of the effectively transmitted flow. And for this, these models consider the activity variations of the requested video sequences, the interaction between the sessions accepted in the VoD system and the variat ion of the capacity in the communication channel. Besides, these models allow us to calculate different statistic measures of user-level QoS. Moreover, these MRC models have a clear understanding with the design and operation of the VoD system. Thus , new design options can be proposed and be easily evaluated. The modelling methodology has been applied to evaluate a real system (testbed) of a video-streaming service. Numerical results obtained from the developed proposals have been compared with results obtained from this testbed, showing a good level of accuracy. This validates the developed approach for a VoD service and, thus, the applicability of the proposed generic method. Then, to reduce the number of states of the VoD system, the designed model has been modified. This way, two new models have been obtained, which produce substantial reductions in the run-time of their evaluation. In addition, models with a lower number of states have been proposed, which produce bounding values of the measures under evaluation. Also, we have carried out a brief study of the most proper SLA parameters to evaluate video services. We have proposed new SLA parameters and new SLA templates. Finally, as a contribution to the methods to solve MRCs, new solutions have been proposed to a mathematical problem very related with the evaluation of MRCs. This mathematical problem is known as the Probability Distribution Function (PDF) of the Linear Combination of Order Statistics (LCOS) of n random variables, independent and uniformly distributed in [0, t]. These solutions presented are numerically stable and their graphical and algebraic interpretations are very simple. Based on modelling and evaluation of adaptive VoD system with different configuration options, we conclude that: the methodology to model adaptive VoD systems which has been proposed in this Doctoral Thesis helps to design and evaluate different VoD systems. Specifically, our proposal helps to design different policies of admission control, different billing methods, and different user profiles, among others. Furthermore, all these design options can be evaluated under different marketing and financial objectives, different incentives to motivate the requirement of the service, different charging methods and so forth.