Database Reference
In-Depth Information
caching approach by preventing violations, reducing storage usage, and enhanc-
ing network resources provision. Furthermore, our replication scheme is easy to
adopt and flexible enough to offer interoperability with de facto approaches,
including HTTP adaptive streaming technique and BitTorrent protocol [
32
].
This work is organized as follows. In Sect.
2
, we present the context and chal-
lenges of this research. We describe in details our prediction model to rank Inter-
net videos in order of demand in Sect.
3
. Section
4
describes the approach of our
adaptive replication scheme, WiseReplica. We explain our simulation method-
ology in Sect.
5
. We then analyse the performance of WiseReplica in Sect.
6
.
Related works are discussed in Sect.
7
, just before the conclusion in Sect.
8
.
2 Context and Challenges
In this section, we describe the context and the challenges of this work.
2.1
Improving Content Availability to Better User Experience
Many studies have shown that quality of user experience while watching online
videos is related to the good quality of content transmission. They presented
many strategies to enhance the video content availability and its distribution.
Most of these studies analyse in the field are focused on Youtube, being this
the major player of video content distribution [
1
,
5
,
6
,
14
]. These studies include
the analysis of crawled data from Youtube APIs and comparisons of caching
strategies from collected data of users' point of view (HTTP logs from ISP or
local networks).
Dobrian
et al.
[
13
] study has shown the correlation between the user engage-
ment and the video quality, being the
Buffering Ratio
(fraction of the total
session time spent in buffering) and
Rendering Rate
(frames per second) the
most critical metric over the total played time for short videos, the current tar-
get of our method. This characterizes the relation between quality of service and
user experience and endures the importance of avoidance of SLAs violations,
which minimizes buffering ratio, confirming the main metric of evaluation in our
work.
Furthermore, Finamore
et al.
[
15
] stated that the download bitrate of the
video plays a fundamental role in video playback quality. They measured the
smoothness of the playback by the bitrate ratio, defined as the ratio between
the average session download bitrate and the video encoding bitrate. Considering
different bitrates in the input dataset is a key aspect of the rendering rate metric
in user engagement and playback quality.
Another important concern in studying the availability of Internet videos was
to provide a SLA-based solution with the minimum constrains regarding deploy-
ment. However, recent studies [
2
,
34
] are based in substantial changes in the
normally used stack of protocols and network infra-structure and they become
hard to be considered as a feasible solution. Our solution takes in account a
Search WWH ::
Custom Search