基于时延Max-Min相似的多路径公平带宽分配

摘要

多路径传输可以提高网络性能,同时也使资源分配的公平性问题复杂化,单路径下能够良好工作的诸多机制无法轻易移植到多路径环境中.根据带宽分配和排队分组、排队时延的内在联系,提出了时延相似性的概念.设计了一种端到端分布式带宽分配算法,将流和传输路径两个层面的控制相分离,动态调整流的排队分组在其各传输路径上的分配,达到相似排队时延,进而实现公平带宽分配.在单瓶颈限定下证明了相似排队时延为公平带宽分配的充分条件.最后,通过两组仿真实验验证了算法的有效性.%Multi-path transmission complicates the problem of fair resource allocation while improving network performance.Few techniques working well in single-path conditions can be smoothly transplanted to a multipath environment.Perceiving the close interactions between queuing packets,delay and bandwidth allocation,this paper proposed a concept of Max-Min delay similarity and based on the concept designed an algorithm for distributed end-to-end bandwidth allocation.By separating control on flow level from on path level,and dynamically reallocating a flow's queuing packets on its transmission path,the network could achieve fair bandwidth allocation by achieving similar delays.It validated the effectiveness of the proposed algorithm by two groups of numeric simulation experiments.Also,given all flows suffer no more than one bottleneck on each path,it proves that delay similarity is a sufficient condition for implementing fair bandwidth allocation.