ITR: Support for Video Traffic in Ad Hoc Networks

Ad-hoc网络是一种新的网络范例，其中网络节点“按需”创建网络。与ad-hoc网络方法相关的特征属性有很多：高度动态的网络体系结构（节点经常加入和离开网络，而且没有任何警告）、完全分布式的体系结构和多跳路由。由于ad-hoc网络体系结构的高重新配置率，许多（如果不是大多数的话）传统路由协议在这种类型的环境中不能很好地执行。因此，提出了一些新的路由协议，特别适合于自组织网络。一个新的互联网工程任务组工作组——MANET工作组已经成立，以解决自组织网络中的路由问题。在ad-hoc网络中的路由上下文中，没有充分讨论的一个主题是多媒体应用程序的qos路由问题，特别是实时流量的路由问题。更具体地说，ad-hoc网络可能会经历频繁且不可预测的网络拓扑变化，这导致网络路径的生命周期相对较短。因此，路径经常变得无效，而且更严重的问题是，可能很少有关于路径向下的警告。尽管对于非实时应用程序来说，这可能不是一个实质性的问题，但这种频繁的路径失效通常会导致实时通信的严重退化。因此，需要一种机制来补偿ad-hoc网络的这种行为。以前在技术文献中已经提出了许多方法。例如，维护一个辅助路由，这样当主路由发生故障时，系统可以在检测到故障后立即恢复到辅助路由上，这已经被证明是一个很好的策略。然而，由于网络中路径的状态通常是高度相关的，因此在很多情况下，备用路由机制是不够的。因此，主路径的故障通常意味着辅助路径也可能不可用。此外，转换时间可能持续太长，从而导致在此期间信号质量的明显下降。ad-hoc网络的一个特点是在源和目的之间有许多路径。因此，利用这些众多路径的机制必然会比上面的双路径方法表现得更好（例如，在支持实时流量的QoS方面）。此外，与其在任何时候选择一条路径用于特定的连接，更好的方案是始终将信息分布在多条路径之间，可能在各个路径上的信息之间存在一定的相关性，以防止路径的某些子集出现故障。因此，提出的机制包括四个步骤：1)发现源节点和目标节点之间的多条路径，并评估路径可用性的相关性；ii)根据一些“优度”度量（如路径的预期可用性、路径的容量、路径的延迟和抖动等），选择路径子集到活动路径集（APS）中，并制定一个方案，允许对网络路径的这些度量进行评估；Iii)在路径之间编码和传播信息的方法（包括将路径与交通类型的具体要求相匹配）；iv)监控APS路径，估计其QoS参数，并根据路径状态及其相关性更新APS的方案。上述方法是通用的，可以应用于各种实时流量类型。然而，为了使研究更加真实，研究人员选择视频通信作为实时测试应用。因此，研究人员打算在视频通信的背景下提出一套特定的算法/协议，以解决上述四个步骤。例如，他们将确定在不稳定路径上传输压缩视频流量的相关参数，并提出将视频源编码为可在多条路径上传播的多个相关描述的方案。研究人员打算将上述视频流量的多路径传输方案集成到ad-hoc网络环境的综合模拟中，该模拟将包括无线电传播模型、节点移动模型、MAC协议和路由算法（用于发现网络路径）。研究人员将从模拟中收集性能指标，这将使他们能够根据所使用模型的参数确定应用程序级别的视频质量。研究人员希望能够回答各种问题，从在自组织网络环境中支持实时流量的可行性这一非常基本的问题，到哪种类型的路由协议最适合实时流量。该研究将通过分析和模拟工具进行。特别是，在模拟中，研究人员将使用先进的模型来预测无线电传播环境、用户移动性模型和流量生成模型。

Ad-hoc networks are a new networking paradigm, in which the network nodes create a network "on demand." There is a number of characteristic attributes associated with the ad-hoc networking approach: highly dynamic network architecture (nodes join and leave the network often and without warning), totally distributed architecture, and multi-hop routing. Due to the high reconfiguration rate of the ad-hoc network architectures, many, if not most of the conventional routing protocols do not perform well in this type of environment. Consequently, a number of novel routing protocols, specifically suited for the ad-hoc networks, were proposed. A new Internet Engineering Task Force working group, the MANET Working Group, has been established to address the issues of routing in ad-hoc networks. One topic that has not been adequately covered in the context of routing in ad-hoc networks is the issue of QoS-routing for multimedia applications, in general, and the issue of routing for real-time traffic, in particular. More specifically, an ad-hoc network may undergo frequent and unpredictable changes in the network topology, which results in relatively short lifetime of the network paths. Thus, paths become frequently invalid, and, what is more of a problem, there may be little warning of a path going down. Although this might not be a substantial problem for non real-time applications, such frequent path invalidation will most often lead to severe degradation of real-time communication. Thus, a mechanism is needed that will compensate for this behavior of ad-hoc networks. A number of approaches have been previously proposed in the technical literature. For instance, maintaining a secondary route, so that when the primary route fails, the system can fall back onto the secondary route as soon as the failure is detected, has proven a good strategy. However, the secondary route mechanism is insufficient in many cases, as the state of paths in the network is usually highly correlated. Thus, failure of the primary path usually means that the secondary path may not be available as well. Also, the change-over time may last too long , so as to cause a perceptible degradation of the signal quality during this period. One characteristic of the ad-hoc networks is that there are many paths between a source and a destination. Thus, a mechanism that takes advantage of these multitude of paths is bound to perform better (i.e., in supporting QoS for real-time traffic) than the above two-path approach. Moreover, rather than selecting a single path at any time to use for a specific connection, a better scheme would be to always distribute the information among multiple paths, possibly with some correlation between the information on the various paths, so as to protect against failure of some subset of the paths. The proposed mechanism thus consists of four steps: i) discovery of multiple paths between the source and the destination nodes and evaluation of the correlation in the paths' availability; ii) selection of a subset of the paths into an Active Path Set (APS), based on some "goodness'' measures (such as the expected availability of the path, the capacity of the path, the delay and jitter of the path, etc), and a scheme that allows to evaluate these measures for the network paths; iii) a method of coding and spreading the information among the paths(including matching the paths with the specific requirements of a traffic type); iv) a scheme to monitor the APS paths, estimate their QoS parameters, and update the APS based on the state of the paths and their correlation. The above approach is general and can be applied to a variety of real-time traffic types. However, to make the study more realistic, the researchers chose video communication as the real-time test application. Thus, the researchers intend to propose a specific set of algorithms/protocols that addresses the four steps as outlined above, in the context of video communication. For instance, they will determine what are the parameters relevant to transmission of compressed video traffic over unstable paths and propose schemes to code video source into multiple correlated descriptions that can be spread over multiple paths. The researchers intend to integrate the above multi-path transport scheme for video traffic into a comprehensive simulation of the ad-hoc networking environment, that will include a radio propagation model, nodal mobility model, MAC protocol, and a routing algorithm (to discover the network paths). The researchers will gather performance measures from the simulation that will allow them to determine the quality of video at the application level depending on the parameters of the models used. The researchers expect to be able to answer questions, ranging from the very basic issue of viability of supporting real-time traffic in an ad-hoc networking environment to what type of routing protocol is most suitable for real-time traffic. The research will be performed through both analytical and simulation tools. In particular, for the simulation, the researchers will use advanced models for prediction of the radio propagation environment, the user mobility model, and the traffic generation model.