DiffServ域下视频流评估关键因素分析

126 浏览量更新于2024-08-29 收藏 247KB PDF 举报

随着图像处理和网络技术的快速发展，视频应用变得日益普及。在这样的背景下，DiffServ（区别服务）架构因其能够为同时传输的视频和数据流提供服务质量保障而受到广泛关注。本文主要针对《国际信号处理、图像处理与模式识别》杂志2015年第8卷第8期的一篇研究论文——"评估DiffServ域上的视频流"进行深入解析。该论文由作者郑万撰写，他隶属于江西财经大学信息技术学院，通信地址位于中国南昌。他的电子邮件地址为wanzheng97@163.com。论文的DOI为http://dx.doi.org/10.14257/ijsip.2015.8.8.06，ISSN为2005-4254，发表在IJSIP（国际期刊《信号处理、图像处理与模式识别》）上，并获得了2015年SERSC版权。文章的核心内容聚焦于对DiffServ域中视频流性能的评估，利用了ns2模拟器中的myEvalvid扩展工具。作者通过全面的研究得出以下关键发现： 1. **多级丢弃优先级策略并非必需**：研究发现，在DiffServ的设置中，使用TSW2C或多级丢弃策略并不一定能显著提升视频流的性能。这表明，简单的策略设计可能已经足够满足大部分场景的需求。 2. **调度模式影响显著**：视频流的传输质量受调度模式的影响很大。例如，在带权轮询（WRR）或带权间隔轮询（WIRR）模式下，参数权重的选择至关重要；而在最高消费带宽（PR）模式中，“最大消耗带宽”参数的设定直接影响到视频流的稳定性和流畅性。 3. **RED参数影响较小**：随机 Early Detection (RED) 是一种流量整形机制，但论文指出其对视频流的影响相对较小，可能在某些特定情况下对服务质量的影响不如其他因素明显。总结来说，这篇研究提供了深入理解DiffServ在网络中如何优化视频流性能的关键见解，对于网络工程师和研究人员来说，它强调了在设计和优化服务质量时应关注的核心策略和参数选择。通过这些发现，实践者可以更有效地在DiffServ环境中部署和管理视频流，确保用户体验的质量。

International Journal of Signal Processing, Image Processing and Pattern Recognition

Vol.8, No.8 (2015)

2.3. Simulation Topology

Figure 1 presents the simulation topology. S1 and S2 generate a video stream and a

CBR data stream respectively. E1, E2 and C forward packets for the sources. Packet size

is 1500 bytes. Bandwidth of each link is set as the figure shows.

C E2 D

DiffServ Domain

5M 2M 5M

Figure 1. Simulation Topology

2.4. Video Sequences

In most experiments, sequence news with CIF resolution is employed. Sequence

foreman and akiyo with CIF resolution are also adopted in particular experiment. Table 1

shows data rate at each second of the three sequences.

Table 1. Data Rate at Each Second of Three Sequences (kbps)

Second No.

news

foreman

akiyo

1280.94

2105.10

721.42

1086.26

1901.77

668.98

1228.94

2002.18

700.13

1414.52

2011.04

761.26

1205.35

2120.10

639.94

1184.26

2172.62

624.94

1326.05

2470.14

692.36

1171.31

2714.23

695.72

1232.28

3153.16

663.18

1372.17

3332.78

773.25

3. Evaluation

According to the type of policy model, experiments are classified into two categories.

Some experiments employ NullPolicer, and the other use TSW2CMPolicer.

NullPolicer does not downgrade any packets. We use this policer to evaluate the

interaction between video and data streams, and discuss the influences of employing

different parameter settings. Since there is no downgrade queue, each physical queue has

only one virtual queue.

3.1. Interaction between Two Streams

To evaluate the interaction between video and data streams, we adopt the default

scheduling mode (RR), the default dropping mode (RIO-C) and the default RED

parameters. The lower and the higher queue length thresholds are 5 and 15, and the

dropping probability is 0. That is to say, RED mechanism is not active.

In the first experiment, the relationship between video receiving quality and generating

rate of data stream (R

) is investigated. Figure 2 presents the results, including the number

of received packets (pktNum) and average PSNR (avgPSNR). If the video stream (news)

is perfectly received, there must be 1690 packets and the corresponding average PSNR

must be 43.987173, as the results when R

equals to 0.6/0.65/0.7Mbps show. When R

increased to 0.75Mbps, packet loss occurs. With the continuous increase of R

, more and

more packets are dropped, resulting in low average PSNR. To give explanation of the

results, let’s recall from figure 1 that the bottleneck bandwidth of the network is 2Mbps,

locating in the link from node C to node E2. From table 1 we can find that the data rate of

剩余11页未读，继续阅读

weixin_38667697

粉丝: 10
资源: 913

DiffServ域下视频流评估关键因素分析

DiffServ域中基于优先级的视频流优化策略

基于DiffServ网络的流视频分发系统优化

DiffServ仿真工具TCL脚本介绍及使用

DiffServ仿真

DiffServ4NS-开源

MPLS DiffServ-aware流量工程

sim.rar_DiffServ tcl

DiffServ队列调度算法研究.pdf

DiffServ队列调度算法优化与性能评估

提高音频视频服务优先级：IntServ与DiffServ协议详解

最新资源