基于效用距离的分层多播反馈控制算法

源自适应的分层多播存在着因反馈信息爆炸而导致的缩放性问题，反馈信息合并是一个较好的解决办法。本文给出一个以效用最优为目标的自适应分层多播模型UALM，分析了在反馈树的支持下，通过反馈合并策略实现反馈控制的方法，提出了一种基于加权效用距离的分层多播反馈合并算法。仿真结果表明，该算法能够有效地控制反馈信息数目，同时使得在此基础上完成的动态分层效用接近于不进行反馈合并时的系统效用，满足分层多播的缩放性要求。

A feedback merging algorithm based on utility distance in adaptive layered multicast

Feedback merging mechanism has been proposed as an effective solution to the feedback implosion problem which arises in large scale source-adaptive layered multicast applications. In this paper, a multicast model named UALM (utility-oriented adaptive layered multicast) was presented firstly, in which utility was introduced as the metric to measure the users satisfaction with the applications. Then the feedback merging mechanism was elaborated to control the amount of feedback messages while minimizing the loss of utility due to feedback merging. In particular, different merging rules were examined in detail and a merging algorithm based on weighted utility distance was presented subsequently. Simulation results showed that the merging algorithm led to much less feedback arrivals at the senders side while maintaining the system utility comparable to that without feedback merging mechanism. It suggests that UALM can be applied in a scalable manner when coupled with this algorithm.