|
In recent years, various solutions offering video-on-demand (VOD) services based on a hierarchical network structure have been proposed.Although many issues have been addressed and studied independently, very few of them discuss the problem of efficient VOD configuration planning in a distributed scenario. In this thesis, we propose a procedure that allows efficient planning of VOD configuration and user load assignment scheme in a distributed scenario. For the problem of configuration planning, we wish to minimize the cost of the configuration while handling a given forecasted load and system blocking probability. Firstly, we analyze the user's traffic behavior by using queueing theory to find the required number of video servers. Then, we employ an optimization technique for solving the so-called Apportionment Problem to determine the optimal number of copies per film. This technique is borrowed from the theory of resource allocation problem, which arises when it is asked to find the optimal allocation of copies in a proportional system. Finally, we model the problem of allocating films to video servers as a bipartite graph and formulate the problem as a linear and integer programming problem that makes a good quality assignment of films to video servers. For the problem of user load assignment, we focus on the of a good assignment scheme such that most of the user requests can be gratified and we want to minimize the cost in a fixed hardware configuration. It handles the on-line assignment of user-to-server, based on the results of the film-to-server assignment and uses an optimization technique to determine the optimal load assignment at any given moment.
|