【摘 要】 A key challenge in the design of parallel video server is ensuring scalability and load balancing. The placement of movie files has a significant effect on these issues. In this thesis, the objective is to find a file allocation algorithm that can di... 更多 >> A key challenge in the design of parallel video server is ensuring scalability and load balancing. The placement of movie files has a significant effect on these issues. In this thesis, the objective is to find a file allocation algorithm that can distribute video files among the video servers in a parallel video server array so as to minimize the load imbalance level and also the storage used. We formulate the movie allocation as a bi-objective function and propose an integrated Genetic Algorithm with Hill Climbing Algorithm to find good allocation schemes for video files. To handle the problem of genetic drift that usually occurs in genetic algorithms, an archive of a Pareto-optimal set is maintained that serves as a pool of possible parents in the selection process of the genetic algorithm. The experimental results show that our algorithm can produce an allocation scheme that balances the load over the server array while using small storage size. << 收起
