Maximizing Streaming Flows Based on a Novel
Video Streaming Framework
Tian W ang
College of Computer Science
and Technology
Huaqiao University
Xiamen, Fujian Province
Email: wsnman@gmail.com
Weijia Jia
Department of Computer Science
City University of Hong Kong
83 Tat Chee Avenue, Kowloon, Hong Kong
Email: itjia@cityu.edu.hk
Bineng Zhong
College of Computer Science
and Technology
Huaqiao University
Xiamen, Fujian Province
Email: bnzhong@gmail.com
Abstract—Video streaming is a kind of bandwidth hungry
application. As a consequence, the number of streaming flows
may be restricted. In this paper, a novel video streaming
framework is designed, where multiple NVSs (Network Video
Servers) form into a server group to collaboratively provide
quality services. A novel problem – Maximum Streaming Flows
(MSF), aiming to maximize the number of simultaneously online
users is proposed. This problem is proved to be NP-Complete
and can be simplified to MSF-2 by adding relays restriction. We
design a (1 − 𝜖) approximation algorithm, where 𝜖 is a constant
which tends to be infinitesimal with the increasing number of
successful streamed flows. We conduct extensive simulations to
show the effectiveness of the methods proposed as compared with
several traditional solutions.
I. INTRODUCTION
Video streaming consumes lots of bandwidth compared with
other data such as audio or web data. There are mainly two
general architectures for real-time video streaming system.
The first one is client-server (C/S) scheme where the data
sources stream the data captured to a server first and remote
users then retrieve the video from the server. This scheme is
simple but incurs a heavy burden on the servers and does not
scale well. By contrast, Peer-to-Peer (P2P) streaming scheme
greatly alleviates the burdens of servers with good scalability.
However, one of the problems about P2P scheme is that the
video sources act as both clients and servers which incurs
burden for sources such as IP cameras. Moreover, it is known
that media streaming over best-effort packet networks such as
the Internet or wireless networks is quite challenging because
of some of the dynamic and unpredictable factors such as
available bandwidth, loss rate, and delay [1]. If the current
connection is unstable, the media flows cannot be streamed to
users or the QoS is not acceptable.
To address these problems, we proposed a novel video
streaming framework. The basic idea is that several video
servers cooperatively form a server group to provide both
streaming and storage services. Different from the traditional
C/S architecture, multiple servers cooperatively undertake the
streaming task. The proposed new streaming scheme can fully
utilize the “server diversity” of dynamic networks. It has
been shown in [2] that usage of multiple streaming servers
provides better robustness in case one of the channels becomes
congested. We expect that multiple servers and path diversity
help in achieving higher overall throughput to the end users.
We mainly make the following contributions: We propose
a novel video streaming framework which introduces multiple
video servers to collaboratively provide services for end users.
Based on this framework, we formulate a novel problem:
Maximum Streaming Flows (MSF) , which aims to maximize
the total number of simultaneous online users. This problem
is proved to be NP-complete. For applications in reality, we
add two-hops relay restriction and get the simplified MSF-2
problem. We design a approximation algorithm with a provable
performance bound compared with the optimal solution.
II. R
ELATED WORK
Video streaming aims at providing high quality video con-
tent to users of both live and on-demand services. Traditional
client-server based video streaming solutions incur expensive
bandwidth provision cost on the server and are not scale
well [3]. On the other hand, Peer-to-Peer (P2P) streaming
greatly alleviates the burdens of servers with good scalabil-
ity [4]. In these typical P2P streaming systems, users named
as peers act as both clients and servers, which incurs burden
for sources such as IP cameras. Recently, cloud computing is
redefining the way many Internet services are operated and
provided, including video streaming. Paper [5] proposes a
predictive cloud system that dynamically books the minimum
bandwidth resources from multiple data centers for the VoD
provider. Our proposed framework is also a cloud computing
streaming architecture and can dynamically exploit bandwidth
resources.
The streaming framework proposed in this paper also takes
advantage of the “server diversity” [2]because original video
sources can pick out better servers to act as steaming servers.
Our work is different from [2] where one source splits
streaming media into several sub-streams and each sub-stream
is delivered separately to the individual users. In our work
we assume one flow is streamed through an exclusive path
which reduces the complexity of implementation. It makes
our problem be different from other problems such as Multi-
commodity problem and traditional routing algorithm [6]
cannot be used.
Proceedings of the 2013 25th International Teletraffic Congress (ITC)
978-0-9836283-7-8 © 2013 ITC
Poster Paper