THESIS
2014
ix, 33 pages : illustrations ; 30 cm
Abstract
With
the
rapid
growth
of
video
streaming
services,
video
content
provider
and
network
service
provider
become
interest
in
the
performance
of
streaming
video
on
their
web
sites
or
networks,
since
video
and
network
quality
dynamically
influence
the
Quality
of
Experience
(QoE).
This
thesis
presents
the
first
large-scale
study
characterizing
the
impact
of
video
and
Wi-Fi
network
performance
on
the
engagement
of
video
viewing
from
the
perspective
of
both
video
content
provider
and
network
service
provider.
Our
study
last
three
months
long
and
collect
7570
anonymized
video
sessions
from
more
than
50
off-the-shelf
routers
in
campus
of
Nanjing
University
of
Posts
and
Telecommunications,
China....[
Read more ]
With
the
rapid
growth
of
video
streaming
services,
video
content
provider
and
network
service
provider
become
interest
in
the
performance
of
streaming
video
on
their
web
sites
or
networks,
since
video
and
network
quality
dynamically
influence
the
Quality
of
Experience
(QoE).
This
thesis
presents
the
first
large-scale
study
characterizing
the
impact
of
video
and
Wi-Fi
network
performance
on
the
engagement
of
video
viewing
from
the
perspective
of
both
video
content
provider
and
network
service
provider.
Our
study
last
three
months
long
and
collect
7570
anonymized
video
sessions
from
more
than
50
off-the-shelf
routers
in
campus
of
Nanjing
University
of
Posts
and
Telecommunications,
China.
Our
study
makes
three
contributions.
First,
we
identify
different
features
on
video
and
Wi-Fi
network
system
that
can
effect
the
engagement
of
video
viewing.
Our
results
can
provide
guidance
to
video
content
provider
and
network
service
provider
on
how
to
improve
the
engagement
directly.
For
example,
give
a
rate
adaptation
based
on
the
change
of
data
rate
and
data
rate
variance.
Second,
we
model
the
complex
relationships
between
video
and
network
factors
and
engagement
of
video
viewing.
Our
model
can
predict
the
percentage
of
video
viewed
with
only
10%
error.
Moreover,
we
generate
a
pruned
model
that
can
use
the
features
obtained
at
the
initial
period
to
predict
engagement,
which
can
be
deployed
as
a
monitor
for
video
content
provider
and
network
service
provider
to
detect
the
performance
of
video
and
network
quality.
We
believe
that
our
model
can
give
some
insight
for
both
video
service
providers
and
network
operators
for
user
engagement
improvement.
Post a Comment