一类混合资源定位服务的优化模型

对一类用于分布式VOD系统的混合了单步k随机漫步和全局中心索引的资源定位服务,用基于Markov过程的模型来描述其中的定位过程,并引入可以描述额外信息的事件,使得其控制不但依赖于状态,而且还依赖于事件.在此基础上,给出了可以直接处理事件的策略优化方法以及状态聚集的相关问题.最后,通过实例仿真验证了所给出的模型和方法.

The Optimizing Model for a Class of Hybrid Resource Location Service

A class of hybrid resource location service for distributed VOD system is studied.It combines one-hop k-random walk and global centralized indexing service.First,this paper proposes a Markov Processes-based model to describe the process of location.Instead of extending the state space,it adopts control related with not only states but also events,which introduce additional information as control basis.Then,a policy iteration algorithm,which can directly deal with the event-related policy optimization,and the aggregation of states according to some particular features,are also introduced.Finally,an example follows to illustrate the proposed model and methods.