| Summary: | 碩士 === 逢甲大學 === 資訊工程所 === 92 === This paper proposes an optimization approach to designing a video-on-demand (VOD) system with the following characteristics: 1) distributed server architecture, 2) heterogeneous system, and 3) a large number of videos. There are three objectives in this optimization design: 1) minimizing the storage cost, 2) minimizing the blocking probability, and 3) balancing the storage load of servers. The video placement is optimized by replicating video. The investigated problem is formulated as a large multi-objective parameter optimization problem with constraints in essence, which is an NP-hard problem. We use a newly-developed intelligent multi-objective evolutionary algorithm (IMOEA) which can efficiently solve large multi-objective parameter optimization problems to obtain a set of high-quality non-dominated solutions. The non-dominated solutions can provide decision makers more flexible solutions to set up appropriate VOD systems according to their preference. If there are some factors of environment changed when the system is operating, IMOEA can adjust video placement quickly and efficiently using its inheritance ability. Experimental results demonstrate the effectiveness of this approach and show that IMOEA performs better than the existing multi-objective algorithm SPEA2 in solving the optimization problem.
|
|---|
