Goodput优化：多路径实时流量负载分配策略

127 浏览量更新于2024-08-26 收藏 1.17MB PDF 举报

"本文主要探讨了在多路径网络上实时流量的负载分配问题，提出了一种名为Goodput-Aware负载分配TiON（GALTON）的模型，旨在优化吞吐量并减少延迟和数据包丢失对流量性能的影响。GALTON模型包含三个阶段：路径状态估计、流量分配和分组交织以满足最后期限约束。该模型基于效用理论进行了数学建模，并通过Exata半物理仿真进行了验证，显示在吞吐量、视频质量、延迟和丢包率等方面优于传统流量分配方法。" 在多路径网络环境中，由于网络路径的多样性和不可靠性，实时流量的性能往往受到较大影响，如端到端延迟增加和数据包连续丢失，这些都会显著降低流量的Goodput，即实际可用的数据传输速率。传统的研究通常侧重于延迟或吞吐量的优化，但GALTON模型则以Goodput为中心，尝试解决这个问题。 GALTON模型的三个阶段如下： 1. 路径状态估计：此阶段的目标是准确地监测和评估每个传输路径的质量，包括带宽、丢包率和延迟等关键指标。这有助于理解网络环境的变化，为后续的流量分配提供依据。 2. 流量分配：根据路径状态估计的结果，GALTON模型将实时流量智能地分配到不同的路径，以最大化总体吞吐量。这一过程涉及到复杂的优化算法，旨在在保证服务质量的同时，充分利用网络资源。 3. 最后期限约束的分组交织：在这一阶段，模型考虑了数据包的到达时间，通过适当的交织策略，防止连续的数据包丢失，从而改善实时流量的稳定性。为了量化GALTON模型的效果，研究人员使用了Exata仿真平台，其中包含了实际的Internet流量追踪和H.264视频流，这使得实验结果更具现实意义。实验结果显示，GALTON模型在多个性能指标上都表现出色，包括提高了吞吐量，提升了视频的PSNR（峰值信噪比），降低了端到端延迟，并减少了总丢失率。 GALTON模型为多路径网络上的实时流量管理提供了一种有效的方法，通过全面考虑网络条件和流量特性，实现了更优的性能。这一研究对于网络资源的管理和优化，特别是在多媒体通信、物联网和云计算等领域，具有重要的实践价值和理论指导意义。

http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI

10.1109/TPDS.2014.2347031, IEEE Transactions on Parallel and Distributed Systems

IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS, VOL. XX, NO. X, 2014 3

the ﬂow rate assignment. An opportunistic scheduling-

based scheduler was proposed to send packets to multi-

ple paths while keeping the fraction of bytes transmitted

on each path.

The ﬂow-based trafﬁc splitting models address the

packet reordering problem by assigning the packets be-

long to the same ﬂow onto the same path. Although the

risk of packet out-of-order may be decreased, load imbal-

ance among the communication paths frequently occurs

and the queueing delay over a single path signiﬁcantly

increases. The Load Balancing for Parallel Forwarding

(LBPF) [15] model selects the path for a ﬂow according to

the hashed result of the packet identiﬁer in the ordinary

mode, similar to the schedule of Direct Hashing. The

FlowLet Aware Routing Engine (FLARE) [16] switch-

es packet bursts named ﬂowlets onto available paths.

Flowlets are spaced by a minimum interval δ, chosen

to be larger than the delay difference between parallel

paths to reduce packet reordering.

The MultiPath LOss Tolerant (MPLOT) [19] transport

protocol addresses the burst data losses in wireless net-

works by introducing a hybrid proactive/reactive FEC

(Forward Error Correction) mechanism. The Encoded

Multipath Streaming (EMS) [20] model adjusts the FEC

redundancy in a progressive manner based on the infor-

mation loss rate. This study reveals the tradeoff between

loss and delay performance in real-time applications.

The work closest to ours is conducted by Bui et al. [2].

The authors formulate the multipath data distribution as

a Markov Decision Process and propose an Online Policy

Iteration (OPI) algorithm. The proposed algorithm aims

at improving the performance metrics, e.g., delay, loss,

and throughput. However, goodput is able to better indi-

cate the QoS requirements of real-time applications. Both

the trafﬁc and network info should be comprehensively

analyzed. Towards this end, we formulate the data distri-

bution aiming to optimize the goodput performance and

develop solutions for ﬂow rate assignment. Besides, we

take into account the burst path losses in heterogeneous

overlay networks and propose an interleaving scheme

to alleviate consecutive losses. In literature [21], we

propose a Sub-Frame Level (SFL) scheduling approach to

optimize the delay performance of high deﬁnition video

streaming over heterogeneous wireless networks.

3 SYSTEM MODEL AND PROBLEM FORMULA-

TION

We ﬁrstly describe the system overview of the proposed

GALTON, which is shown in Fig. 2. The proposed

model is an end-to-end scheduling framework and the

implementation requires the modiﬁcations on the sender

and receiver sides.

The key working component in the sender side is the

parameter control unit, which takes in the input pa-

rameters (including the path status, deadline constraint)

and outputs the scheduling parameters (including ﬂow

assignment vector, interleaving level, and probing trafﬁc

adaption parameters). The ﬂow rate allocator is respon-

sible for partitioning the input trafﬁc (which may contain

multiple deadline-constrained ﬂows) into several sub-

ﬂows and dispatching each of them to the available

paths. The allocated sub-ﬂows will be temporarily stored

in the sending buffers for different communication paths.

In the process of deadline-constrained packet interleav-

ing, the packets scheduled onto the same path will be

spread out within the delay constraint.

The receiving buffer at the receiver side is used for

collecting the packets. In order to restore the original

trafﬁc, packet-to-ﬂow mapping and inter-packet rese-

quence are two necessary steps. The problem of trafﬁc

load distribution involves the models of communication

path and real-time trafﬁc, which will be described in the

rest of this section.

3.1 Communication Path Model

We consider a heterogeneous overlay network integrat-

ing multiple communication paths between two termi-

nals. The end-to-end connection can be constructed by

binding a pair of IP addresses from the source and

destination node, respectively. Each communication path

p ∈ P is considered to be an independent transport

link uncorrelated with others and this assumption can

be justiﬁed as follows

• In the environment of wired multipath networks,

a method commonly used in existing studies (e.g.,

[22][36][46]) is to detect the shared congestion paths

with the end-to-end measurement techniques (e.g.,

the cross-correlation based approaches [47], and

wavelet based schemes [48]). Then, the paths shar-

ing bottleneck links can be treated as one com-

munication path. Another method to construct the

end-to-end independent paths is to use the overlay

relay nodes to forward the trafﬁc [49]. In this case,

the available communication paths are disjoint. The

general review of setting the independent paths can

be found in [50].

• In the context of heterogeneous wireless networks

(e.g., WLAN, Cellular, WiMAX), the last-hop wire-

less links are most likely to be the bottleneck links of

the end-to-end paths due to the limited capacity and

time-varying channel status [6][39]. These access

networks use different wireless spectrums and do

not interfere with each other. Therefore, we can

assume the available capacities and loss rates are

independent to each other [6][7].

The end-to-end communication paths are character-

ized by the following properties:

• the available bandwidth µ

. This metric does not

indicate the raw per-path capacity, but the time-

varying share of that bandwidth as perceived by

the end-to-end ﬂow. It can also be viewed as the

sending rate on path p that is achievable with the

current sending buffer and round trip time.

剩余13页未读，继续阅读

weixin_38591291

粉丝: 6
资源: 957

Goodput优化：多路径实时流量负载分配策略

Throught与Goodput的差别及测试软件教程.rar

异构无线网络中用于移动视频流的失真感知并发多路径传输

链路传输速率为64kbps，单向传播时延为15ms。设数据 帧长为3200bit，确认人帧长度为 128bit， 在误帧率p=0.01情 况下，MATLAB编程程序

卡通风格化魔法术技能粒子特效 ：Toon Projectiles 2 1.0

在 MATLAB GUI 中动态更新数据：策略与实践

【JCR一区级】Matlab实现白鹭群优化算法ESOA-CNN-BiLSTM-Attention的故障诊断算法研究.rar

信创实验室建设方案（24页）.pptx

KGBrowserSetup-x86-V1.0.0.100-20190315.exe

obspy-1.2.2-cp38-cp38-win_amd64.whl

数字政府大数据政务云平台顶层设计方案(75页）.pptx

最新资源

链路传输速率为64kbps，单向传播时延为15ms。设数据帧长为3200bit，确认人帧长度为 128bit，在误帧率p=0.01情况下，MATLAB编程程序

卡通风格化魔法术技能粒子特效：Toon Projectiles 2 1.0