[论文解读] Design and implementation of a differentiated service based qos model for real-time interactive traffic on constrained bandwidth ip networks

Sruti Gan Chaudhuri|arXiv (Cornell University)|May 15, 2012

Network Traffic and Congestion Control参考文献 6被引用 3

一句话总结

本文提出了一种基于DiffServ的QoS模型，用于带宽受限的IP网络中的实时交互式流量，通过基于优先级的调度和加权服务速率实现动态自适应队列，确保低延迟和低丢包率，且无需过度配置资源。实验结果表明，该模型在实验室和真实网络测试中均能有效为视频会议和监控流量提供QoS保障，显著提升了公平性和资源利用率。

ABSTRACT

In this thesis work, a QoS model for real-time interactive traffic on a real network with constrained bandwidth and real-time traffic has been proposed. The model supports tight guarantees of QoS to real-time interactive traffic without over provisioning of bandwidth. A dynamic scheduling model which is adaptive to input data rate of traffic has been proposed. In this model, A Differentiated Service (DiffServ) based approach is proposed for QoS provisioning. The packets are classified and distributed among finite number of queues with limited buffer based on different priorities and total available bandwidth. The model proposes a mechanism to derive the weighted service rates and queue length distribution so as to meet the requirement of low packet loss and delay for real time interactive traffic in the QoS engineered network. An adaptive queuing strategy is proposed so that minimum bandwidth in used for real time traffic. This ensures maximizing availability to best effort traffic. The model assumes constrained bandwidth without having to over provision the network resources and thus keeping the cost low. A modified version suitable for testing on a real network is also presented. Experimental verification of these in a test bed network in a laboratory as well as on a real network has been carried out. The results of the QoS provisioning model for different sources of real-time traffic such as video conferencing equipment, robotic surveillance camera has also been shown. The thesis also introduces a real-time Variable Bit Rate (VBR) traffic tuning parameter for controlling the service of VBR traffic to give better and fair performance to rest of the traffic.

研究动机与目标

• 设计一种QoS模型，确保在带宽受限的IP网络中，实时交互式流量具有低延迟和低丢包率。
• 在维持实时应用高质量服务的同时，避免网络过度配置。
• 开发一种自适应队列策略，能够动态适应输入数据速率并优化带宽使用。
• 在带宽有限的情况下，确保实时流量与尽力而为流量之间的公平资源分配。
• 在实验室测试平台和真实网络环境中对模型进行实验验证。

提出的方法

• 该模型采用区分服务（DiffServ）框架，将数据包分类至基于优先级的队列，并限制缓冲区空间。
• 实施一种动态调度机制，能够根据实时流量的输入速率自适应调整。
• 通过数学推导加权服务速率和队列长度分布，以满足延迟和丢包的QoS要求。
• 自适应队列策略可最小化实时流量的带宽使用，从而最大化为尽力而为流量保留的可用带宽。
• 引入一种实时可变比特率（VBR）流量调参机制，以在不同类型的流量之间实现性能平衡。
• 在实际网络基础设施上，对模型的改进版本进行了真实环境测试。

实验结果

研究问题

• RQ1 如何在不进行网络过度配置的前提下，实现带宽受限IP网络中实时交互式流量的QoS保障？
• RQ2 何种动态调度机制可在输入流量速率变化时，确保低延迟和低丢包率？
• RQ3 如何优化加权服务速率和队列长度分布，以满足实时QoS要求？
• RQ4 VBR调参参数对混合流量类型下公平性和性能的影响如何？
• RQ5 与理论预期相比，该模型在真实网络环境中的表现如何？

主要发现

• 该模型在实验室和真实网络环境中均成功实现了对实时流量（如视频会议和机器人监控）的低丢包率和低延迟。
• 实验结果证实，自适应队列策略有效最小化了实时流量的带宽使用，同时保留了为尽力而为流量服务的容量。
• VBR调参参数实现了更公平的性能分配，显著提升了整体网络效率。
• 该模型在无需网络过度配置的情况下维持了QoS保障，从而降低了运营成本。
• 在真实测试平台网络上的实现验证了该模型在实际条件下的可行性和有效性。
• 用于推导加权服务速率和队列长度分布的机制，有效支持了低延迟、低丢包的运行。