Chunk-Level Request-Grant-Transfer Mode for QoE-Sensitive
Video Delivery in CDN
Gengbiao Shen
1,2
, Qing Li
3,2
, Yong Jiang
1,2
, Richard Sinnott
4
,
Dong Lin
3,2
, Zehua Guo
5,6
, Yi Wang
3,2
1
Graduate School at Shenzhen, Tsinghua University, Shenzhen, China
2
PCL Research Center of Networks and Communications, Peng Cheng Laboratory, Shenzhen, China
3
Southern University of Science and Technology, Shenzhen, China
4
University of Melbourne, Melbourne, Australia
5
Beijing Institute of Technology, Beijing, China
6
University of Minnesota Twin Cities, Minnesota, USA
Emails:sgb16@mails.tsinghua.edu.cn,liq8@sustech.edu.cn,jiangy@sz.tsinghua.edu.cn,rsinnott@unimelb.edu.au
lind@sustech.edu.cn,guolizihao@hotmail.com,wangy37@sustech.edu.cn
ABSTRACT
Remote Direct Memory Access (RDMA) can be deployed
in Content Delivery Networks (CDN) Points of Presence
(PoPs) to avoid the high CPU overheads caused by tradi-
tional TCP/IP stacks. However, RDMA cannot surmount
the drawbacks of the window-based conservative of TCP
and is insensitive to Quality of Experience (QoE). Moreover,
the requirement of lossless networks hinders the widespread
application of RDMA. In this paper, we introduce the paral-
lel multipoint-to-multipoint Request-Grant-Transfer (RGT)
mode into RDMA to solve the aforementioned problems.
Compared with traditional RGT mode, our scheme supports
parallel Dynamic Adaptive Streaming over HTTP (DASH)
chunk delivery, thereby improving throughput and reduc-
ing initial delays. We differentiate the importance of DASH
chunks according to QoE-related properties. In this way, we
reduce the response time of specific DASH chunks. We pro-
vide an efficient approach to select the optimal number of
requests for partially traversing pending requests to reduce
the overheads of Request stages. We perform comprehensive
experiments to demonstrate that our scheme improves the
throughput of CDN PoPs and enhances client QoE.
CCS CONCEPTS
• Networks → Network protocol design
;
Cross-layer
protocols
;
In-network processing
;
Overlay and other
logical network structures.
Corresponding author: Qing Li (liq8@sustech.edu.cn).
Permission to make digital or hard copies of all or part of this work
for personal or classroom use is granted without fee provided that
copies are not made or distributed for profit or commercial advantage
and that copies bear this notice and the full citation on the first
page. Copyrights for components of this work owned by others than
ACM must be honored. Abstracting with credit is permitted. To copy
otherwise, or republish, to post on servers or to redistribute to lists,
requires prior specific permission and/or a fee. Request permissions
from permissions@acm.org.
IWQoS ’19, June 24–25, 2019, Phoenix, AZ, USA
© 2019 Association for Computing Machinery.
ACM ISBN 978-1-4503-6778-3/19/06. . . $15.00
https://doi.org/10.1145/3326285.3329047
KEYWORDS
content delivery network, RDMA, multipoint-to-multipoint,
DASH, QoE
ACM Reference Format:
Gengbiao Shen
1,2
, Qing Li
3,2
, Yong Jiang
1,2
, Richard Sinnott
4
,
and Dong Lin
3,2
, Zehua Guo
5,6
, Yi Wang
3,2
. 2019. Chunk-Level
Request-Grant-Transfer Mode for QoE-Sensitive Video Delivery
in CDN. In IEEE/ACM International Symposium on Quality
of Service (IWQoS ’19), June 24–25, 2019, Phoenix, AZ, USA.
ACM, New York, NY, USA, 10 pages. https://doi.org/10.1145/
3326285.3329047
1 INTRODUCTION
With the popularity of the Video on Demand (VoD) and
live streams, Content Delivery Networks (CDNs) have been
widely deployed by Content Providers (CPs) to supply clients
with great Quality of Experience (QoE). Nowadays, approxi-
mately up to 69% of the Internet traffic derives from video
applications, while two thirds of all video traffic is delivered
by CDNs [
20
]. Many companies, such as Google, Akamai,
Youtube and Netflix, have established their own CDNs to
realize high availablity and low latency through deploying
several geographically distributed Points of Presence (PoPs)
[
2
,
8
,
24
,
25
].CDNs distribute duplicate content to PoPs and
redirect clients to the closest PoP, which means that the
client QoE depends on the performance of PoPs. Typically,
the architecture of PoPs is based on a multilayer structure
that contains several cache proxies and a number of origin
servers [
10
,
18
] as shown in Fig. 1. Cache proxies serve the
arrived content requests and access to origin servers to obtain
the cache miss content. It is non-trivial to design an efficient
data transfer scheme in this frontend-backend architecture.
However, traditional TCP/IP stacks cannot satisfy the
high-throughput and low-latency scenarios in PoPs. They
consume considerable CPU resources in operating systems
which limits throughput and causes the endpoints to become
a bottleneck. Window-based congestion control has a conser-
vative adjustment in TCP and in the high-speed situations
bandwidth resources cannot be fully utilized. Since Dynamic
Adaptive Streaming over HTTP (DASH) [
22
,
23
] has been