Research And Design Of Architecture And Load Balancing For Large-scale Embedded VOD

Television is the indispensable entertainment tool for everyone's daily life. According to application of embedded VOD, it will combine with digital television set-top box. Then it will provide us with not only entertainment, but also study, work and life, which will change people's way of life eventually without a doubt.The VOD system based on embedded terminal-digital television set-top box discussed in the paper, its architecture and load balancing strategy etc. are not mature enough. Although digital television set-top box has complete real-time operating system and CPU to coordinate and control hardware facilities in it, provides simple graphical user interface. But in addition to multimedia and network technology of traditional VOD system to consider, due to less memory resources, weaker ability to calculate, single functional target of embedded client relative to PC, it can't do large amounts of data storage and handling. While it'll cost much money to update embedded system resources, based on this, it proposed higher demand for embedded VOD server than that is based on PC terminal.The paper researches from the architecture and load balancing of large-scale embedded VOD system, firstly indicates the differences from other ordinary broadband applications, as well as the characteristics of embedded VOD system differs from the PC terminal, analyzes the discrepancy between large-scale and small-scale. Through research of the existing architecture of server, such as node-server architecture, the distributed and the virtual architecture, it designs a kind of distributed virtual architecture, to seek some balance between reusability of broadband and bottleneck brought out by control server, to balance VOD application performance between usability and expansibility.Then according to analysis and research for advantages and disadvantages of existing load balancing strategy, combining the characteristics of embedded VOD applications, it proposed a statistical models-based load balancing design for VOD, commenced from random behavior of users, analyzed the implied rules, extracted statistical model, divided the programs into pieces based on it, copied and stored them according to different order probability of diverse segments in diverse programs, and adjust the program stored in the server nodes according to user's interests and order rate regularly, to implement dynamic load balance between server nodes. After test and comparison of the algorithms, the result of simulation verifies its superiority over other algorithms, the mission success rate is higher, even though the number of user request is large it can still keep certain response speed and success rate, thus improving the stability, robustness and expandability of system, possess certain adaptability.