多路径网络上实时流量的Goodput-Aware负载分配资源-CSDN文库

80 浏览量 2021-04-04 18:10:33 上传评论收藏 1.17MB PDF 举报

资源详情

资源评论

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 1

Goodput-Aware Load Distribution for Real-time

Trafﬁc over Multipath Networks

Jiyan Wu, Student Member, IEEE, Chau Yuen, Senior Member, IEEE, Bo Cheng, Yanlei Shang, and

Junliang Chen

Abstract—Load distribution is a key research issue in deploying the limited network resources available to support trafﬁc transmissions.

Developing an effective solution is critical for enhancing trafﬁc performance and network utilization. In this paper, we investigate the

problem of load distribution for real-time trafﬁc over multipath networks. Due to the path diversity and unreliability in heterogeneous

overlay networks, large end-to-end delay and consecutive packet losses can signiﬁcantly degrade the trafﬁc ﬂow’s goodput, whereas

existing studies mainly focus on the delay or throughput performance. To address the challenging problems, we propose a Goodput-

Aware Load distribuTiON (GALTON) model that includes three phases: (1) path status estimation to accurately sense the quality of

each transport link, (2) ﬂow rate assignment to optimize the aggregate goodput of input trafﬁc, and (3) deadline-constrained packet

interleaving to mitigate consecutive losses. We present a mathematical formulation for multipath load distribution and derive the solution

based on utility theory. The performance of the proposed model is evaluated through semi-physical emulations in Exata involving both

real Internet trafﬁc traces and H.264 video streaming. Experimental results show that GALTON outperforms existing trafﬁc distribution

models in terms of goodput, video PSNR (Peak Signal-to-Noise Ratio), end-to-end delay, and aggregate loss rate.

Index Terms—Load Distribution, Goodput, Multipath Networks, Real-time Trafﬁc, Multihoming.

1 INTRODUCTION

He advancements of various network infrastructures

have reached an unprecedent height during the

past few years. The network heterogeneity and high-

degree connectivity to different access medium provide

increased opportunities to establish multiple paths be-

tween end devices [1] as illustrated in Fig. 1. However,

a single communication path is not capable of com-

pletely satisfying the stringent QoS (Quality of Service)

requirements (e.g., delay, bandwidth and reliability) im-

posed by current and emerging real-time multimedia

applications [2][3] (e.g., multi-player online gaming and

live sports program). With the rise of multihomed or

multinetwork clients [4] (e.g., the Mushroom products

[5]), new research trends [6][7] have moved toward-

s simultaneously exploiting these multiple paths for

enhanced transmission reliability and throughput. The

IETF MONAMI6 working group [8], which is focused on

enhancing mobile IPv6 with multihoming support, has

identiﬁed the beneﬁts that multihomed hosts offered to

both individual mobiles and network operators.

The key research issue in utilizing the available path-

s between multihomed communication terminals is to

effectively distribute input trafﬁc load for providing

adequate QoS perceived by end users [1][2][9]. Indeed,

• Jiyan Wu, Bo Cheng, Yanlei Shang, and Junliang Chen are with the

State Key Laboratory of Networking and Switching Technology, Beijing

University of Posts and Telecommunications, Beijing 100876, P. R. China.

• Jiyan Wu and Chau Yuen are with the SUTD-MIT International Design

Center, Singapore University of Technology and Design, 20 Dover Drive,

Singapore 138682.

E-mail: wujiyan@126.com, yuenchau@sutd.edu.sg, {chengbo, shangyl,

chjl}@bupt.edu.cn.

Cellular

Wi-Fi

Media Server

Multi-mode

WiMAX

Load Distributor

Gateway

Multi-mode

Fig. 1. Illustration of multipath connectivity in heteroge-

neous overlay networks. The communications paths can

be established either in the wired domain (e.g., using

the SCTP - Stream Control Transmission Protocol [44])

or in hybrid wireless networks (e.g., cooperative packet

delivery in cellular, Wi-Fi, WiMAX and ad hoc networks).

inefﬁcient load distribution can signiﬁcantly degrade the

trafﬁc performance and network utilization, e.g., load

imbalance, packet reordering, large end-to-end delay, etc.

Therefore, many algorithms [1][2][10][11] have been pro-

posed to optimize the delay or throughput performance.

However, these network-level criteria cannot properly

indicate the beneﬁts of upper-layer applications. For

instance, a live streaming video application cannot ef-

fectively leverage the throughput gains since its stream-

ing rate is typically ﬁxed or bounded by the encoding

schemes. Furthermore, the increased throughput may

lead to larger end-to-end delays, which in turn induce

video quality degradation. Consequently, the load distri-

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 2

bution of real-time trafﬁc to achieve excellent QoS still

remains problematic.

Distinct from previous studies, we present in this

paper a load distribution model to optimize the goodput

[12] performance of real-time trafﬁc over multipath net-

works. Goodput differs from throughput as it represents

the amount of data successfully received by the destina-

tion within the imposed deadline. The path diversity and

unreliability in heterogeneous overlay networks [13], in

concert with the stringent QoS requirements, pose crucial

challenges to achieve the goal. To effectively aggregate

the available capacity of different network paths, we

have to seriously consider: (1) how to guarantee the

input trafﬁc delivering within the delay constraint, and

(2) how to alleviate the burst data losses

frequently en-

countered in wired/wireless packet switching networks.

Motivated by addressing the above issues, we propose

a Goodput-Aware Load distribuTiON model (GALTON)

that includes three phases: (1) path status estimation

to capture the physical characteristics of each transport

link, (2) ﬂow rate assignment to optimize the aggregate

goodput of input trafﬁc, and (3) deadline-constrained

packet interleaving to alleviate consecutive losses. The

detailed descriptions of the proposed solution will be

presented in Section 4.

Speciﬁcally, the contributions of this paper can be

summarized as follows:

• A load distribution model that effectively integrates

the path status estimation, ﬂow rate assignment, and

packet interleaving to optimize the goodput perfor-

mance of real-time trafﬁc over multipath networks.

• A mathematical formulation for load distribution

of multiple deadline-constrained ﬂows over parallel

communication paths to maximize the aggregate

goodput. The utility theory is employed to derive

the solution for ﬂow rate assignment.

• Extensive semi-physical emulations in Exata involv-

ing both real Internet trafﬁc and H.264 video stream-

ing over wired/wireless multipath networks. Exper-

imental results show that: (1) GALTON improves

the goodput by up to 0.25, 0.48, and 0.64 Mbps

compared to the OPI [2], E-DCLD [1], and THR

[45], respectively. (2) GALTON increases the average

video PSNR by up to 6.1, 8.6, and 12.1 dB com-

pared to the OPI, E-DCLD, and THR, respectively.

(3) GALTON reduces the average end-to-end delay

by up to 33.1, 12.3, and 46.5 ms compared to the

OPI, E-DCLD, and THR, respectively. (4) GALTON

mitigates the aggregate loss rates by up to 5.1%,

9.5%, and 10.9% compared to the OPI, E-DCLD, and

THR, respectively. Furthermore, the superiority of

1. In the context of heterogeneous overlay networks, the packet

losses can be classiﬁed into three categories: 1) the losses caused by

congestions due to the bandwidth limitation or buffer overﬂow; 2) the

errors caused by noise or interference in the wireless networks; and

3) the path failure loss or handover loss. In wireless networks, most

packet losses are due to the wireless channel ﬂuctuations or path failure

and not caused by the link congestion.

GALTON over the competing schemes become more

obvious as the number of available paths increases.

The remainder of this paper is structured as follows.

In Section 2, we brieﬂy review and discuss the related

work. The system model and problem formulation are

presented in Section 3. Section 4 describes the solution

procedure of the proposed GALTON. Performance eval-

uation is provided in Section 5 and concluding remarks

are given in Section 6. The basic notations used through-

out this paper are presented in Table 1.

TABLE 1

Basic notations used throughout this paper.

Symbol Deﬁnition

P, E the probability value, expectation value.

P, p the set of available paths, a path element.

F, f the set of trafﬁc ﬂows, a ﬂow element.

P, F the number of available paths, trafﬁc ﬂows.

R, R

the ﬂow rate assignment matrix, an element.

T the delay constraint for the input trafﬁc.

RT T

the round trip time of p.

, ν

the available, residual bandwidth of p.

G/B the Good/Bad state of p.

, π

the stationary probability that p is in B/G state.

the state transition probability of p from B to G.

the transmission loss rate of p.

the effective packet loss rate of ﬂow f over p.

, P kt

the probing trafﬁc rate/size of p.

M, M

the total number of packets, dispatched onto p.

the end-to-end delay of ﬂow f over path p.

Θ the aggregate goodput.

U, U

the system utility matrix, an element.

2 RELATED WORK

Trafﬁc load distribution over multipath networks has

been an active research area in recent years and the

general reviews can be found in [9][14]. The existing

distribution models can be categorized into the ﬂow and

packet based trafﬁc splitting approaches.

The packet based models generally dispatch packets

onto different paths based on the channel status in-

formation, packet delay constraint, etc. Although these

scheduling policies are able to reduce the queueing

delay over a single path, they may also induce serious

packet reordering problems, which in turn result in

large end-to-end delay. The Effective Delay Controlled

Load Distribution (E-DCLD) [1] aims at minimizing the

difference among end-to-end delays of different paths,

thereby reducing packet delay variation and risk of pack-

et reordering. The authors formulate the queueing delay

for each path with a hybrid M/M/1 model and splits

the trafﬁc to minimize the cost of delay variations. In

[18], each path is assigned with a weight associated with

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

评论收藏

内容反馈

多路径网络上实时流量的Goodput-Aware负载分配

评论0

最新资源

多路径网络上实时流量的Goodput-Aware负载分配

评论0

最新资源

相关推荐

Throught与Goodput的差别及测试软件教程.rar

Firewall-Performance-Testing-Use-Cases-in-Lab-and-Production-Networks.pdf

Performance Testing Virtual Firewalls.pdf

SafireReport-08-Jun-2020.pdf

tcp-metrics:一种使用不同方法测量TCP连接的往返时间和带宽的工具

tcp.rar_TCP NS2_ns2_ns2 tcp_ns2+tcp

80211WLAN中吞吐率的分析论文

Veriwave Apps用户使用手册（企业内部教程，绝对独家）

Qt 5实现串口调试助手 （源工程文件、0积分下载）

【SystemVerilog】路科验证V2学习笔记（全600页）.pdf

AutoSAR标准协议4.2.2

光伏-储能并网系统仿真.rar

XCP协议的规范文档

GD32替换STM32注意事项.pdf

NPPJSONViewer.zip

CANoe通过CAPL脚本实现自动测试

蓝牙BLE协议中文版.pdf

VS2015安装证书，JavaScript_ProjectSystem.msi，JavaScript_LanguageService.msi

BaiduOCR.zip

AD20官方中文教程.pdf

电路分析基础第二版PDF电子书免费下载

完整版 Microsoft.ACE.OLEDB.12.0 驱动下载.rar

Tangent免费.rar

七参数坐标转换工具（可在WGS84、北京54、西安80、CGCS2000坐标系中任意两个转换）

CMSIS-DAP使用说明及驱动.rar

Matlab安装MinGW-w64 C/C++ 编译器

qt样式表一键生成（花狗Fdog）

Elsevier期刊word模板.zip

Qt 5实现串口调试助手（源工程文件、0积分下载）