基于IP网络的自适应QoS管理方案研究

目前实时音频、视频多媒体应用已经开始进入IP网络，但还有许多问题没有得到很好地解决，其中一个关键问题是多媒体服务的QoS问题、TCP／IP协议本身只提供一种“Best-effort”级别的服务，对QoS支持很少，“Best-effort”级别的服务往往会导致实时应用出现延迟抖动、分组丢失率高，从而极大的影响了实时应用的运行效果，因此必须研究可行的、高效的基于IP网络的QoS控制机制。IP网络QoS已成为分布式多媒体和网络通信的重要研究热点和难点课题。本文在IEＴＦ Intserv与Diffserv相结合的体系结构基础上，提出了一种适用于IP网络的自适应QoS管理框架，与Intserv或Diffserv模型不同的是，该QoS管理框架引入优先级节和自适应概念，QoS俦权处理采用基于多媒体优先级节的算法，在传输控制上采用了自适应QoS控制算法。它在应用层上完成，因此独立于底层网络协议。本文首先讨论了基本概念和函数，然后提出一种基于优先级节的自适应QoS管理框架和优先级调度流程。提出一种基于IP的自适应QoS协商机制，基本思想是基于RSVP（资源预留协议）。在提出QoS请求时进行QoS映射，然后启动适应性函数和资源管理函数进行协商，直到获得一组在当前资源状况下最佳的QoS指标。最后，本文还讨论了此方案的有效性并在自行开发的实验平台GUT上进行的QoS实验。实验表明应用该方法可以根据网络当前状态自适应地整多媒体实时应用的QoS需求。

An Adaptive QoS Management Mechanism Based on IP Network

At present, real-time audio and video multimedia applicationshave come to IP network, but many problems have not been resolved yet, such as the QoS management of multimedia services. TCP/IP protocol only provides services of “Best-effort” level, lacking of QoS supporting, which always causes latency jittering and packet loss in real-time multimedia application, and greatly affects the quality of real-time applications. IP QoS has become one of the most important and most difficult research subjects of distributed multimedia application and network communication. So it is necessary to find feasible and highly effective QoS management mechanism based on IP network. On the basis of combining models of integrated services (Intserv) and differentiated services (Diffserv) of IETF, the authors put forward an adaptive QoS management framework for IP network. Different from the Intserv or Diffserv model, the adaptive QoS management framework adopts the concepts of multimedia priority session and adaptive QoS controlling. The QoS priority processing algorithm is based on multimedia priority session, and the transmission procedure adopts an adaptive QoS control algorithm. The QoS management works on the application level, so the QoS mechanism is separated with low level network protocols. The authors at first discuss the basic concepts and functions, such as the multimedia priority session, adaptability function, call admission function, and resource management function, and then propose an adaptive QoS management framework and a priority schedule flow based on multimedia priority session. The authors also propose an adaptive QoS negotiation mechanism which, relying on resource reservation protocol (RSVP), carries out QoS mapping, starts the adaptability function while QoS requesting, starts the resource management function while QoS accepting, and continues to negotiate between both sides until a group of QoS values have been achieved. Finally, the article discusses the effectiveness of the QoS management mechanism, and the experiment done on the distributed multimedia experimental platform developed by the authors. The result shows that, with our approach, the QoS requirement can be adjusted adaptively when the network resources changes.