Quality-Adaptive Scheduling for Live Streaming over
Multiple Access Networks
Kristian Evensen1, Tomas Kupka2, Dominik Kaspar1,
Pål Halvorsen1,2, Carsten Griwodz1,2
1Simula Research Laboratory, Norway 2Department of Informatics, University of Oslo, Norway
{kristrev, tomasku, kaspar, paalh, griff}@simula.no
ABSTRACT
Video streaming ranks among the most popular services of-
fered through the Internet today. At the same time, access-
ing the Internet over public WiFi and 3G networks has be-
come part of our everyday lives. However, streaming video
in wireless environments is often subject to frequent periods
of rebuffering and characterized by low picture quality. In
particular, achieving smooth and quality-adaptive streaming
of live video poses a big challenge in mobile scenarios.
Building on the observation that the subjective video ex-
perience on mobile devices decreases when quality changes
are more frequent than every 1 to 2 seconds, we present a
client-side scheduler that retrieves segments of several video
encodings over heterogeneous network interfaces simultane-
ously. By extending the DAVVI streaming platform with
support for multihoming, the proposed scheduler’s perfor-
mance is experimentally evaluated. The results show that
our scheduler reduces the video interruptions and achieves a
higher and more stable average quality over multiple, truly
heterogeneous wireless interfaces.
Categories and Subject Descriptors
C.2.4 [Computer-Communication Networks]: Distributed
Systems - Client/Server
General Terms
Performance
1. INTRODUCTION
Video streaming is rapidly increasing in popularity and
has become one of the dominant services on the Internet
today. For example, the video aggregation site YouTube
streams millions of videos per day, and over 20 hours of
content is uploaded every hour1. In addition, several live
streaming services exist. Microsoft’s SmoothStreaming [13]
1http://www.youtube.com/t/fact sheet
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. To copy otherwise, to
republish, to post on servers or to redistribute to lists, requires prior specific
permission and/or a fee.
NOSSDAV’10, June 2–4, 2010, Amsterdam, The Netherlands.
Copyright 2010 ACM 978-1-4503-0043-8/10/06 ...$10.00.
is for example used to deliver the 2010 Olympic games quasi-
live to millions of people, with only a small time gap between
the actual event and video display.
Streaming high-quality video requires a lot of bandwidth.
For example, the bit rate of H264-compressed 1080p video
is usually around 6 - 8Mbit/s. If a user is connected to the
Internet using a fixed connection, this may not be a prob-
lem, but if the connection is established using a wireless link,
the available bandwidth often becomes a bottleneck. Even
though the maximum bandwidth of most IEEE802.11 stan-
dards (WLAN) exceeds the requirements for 1080p video,
the experienced throughput is often insufficient. Link con-
tention, interference or weak signal reception are some phe-
nomena that cause a significant throughput reduction.
A possible way to alleviate the bandwidth problem caused
by most wireless technologies is to increase the performance
by combining the throughput of multiple network interfaces.
Today, many client terminals are equipped with multiple
interfaces and are often within coverage range of overlapping
access networks. For example, laptops and smart phones can
often connect to both WLAN and 3G networks. Delivering
a video as a sequence of small, independent movie segments
coded in several qualities, as done by Microsoft’s torrent-
like, HTTP-based SmoothStreaming, enables a stream to
adapt to current network fluctuations and potentially allows
a single video segment to be distributed over various paths
for bandwidth aggregation.
Here, application protocol features, such as HTTP’s range
retrieval requests, allow a file to be logically segmented and
to be downloaded over various access networks [5] or from
redundant sources [10]. If a client’s interfaces are connected
to independent networks, the simultaneous use of multiple
wireless links can achieve a total transfer bit rate close to
the sum of all the individual interfaces’ throughput. In ad-
dition, taking advantage of request pipelining [6, 9, 10], the
aggregated throughput can be further optimized for small
segments. However, obtaining a minimal time gap between
the live event and the video display – in other words, in-
creasing the liveness of a video stream – is a challenge that
increases in complexity when scheduling video segments over
network interfaces with heterogeneous characteristics.
This paper introduces an adaptive, pull-based scheduler
that achieves smooth playback by scheduling requests for
video segments of different quality levels over multiple in-
terfaces simultaneously. The application-layer scheduler is
implemented in our experimental testbed as an extension to
the DAVVI [4] video streaming platform, which allows live
multimedia streaming from commodity web servers.
21