QoE
InternetInternet
Throughput
Predictor
Throughput
Predictor
Bitrate
Controller
Bitrate
Controller
HTTPHTTP
GET
Chunk
Buffer
Occupancy
Video Player
BufferBufferBuffer
End User
Bitrate
Throughput
Prediction
Figure 1: Abstract model of DASH players
is empty and cannot render the video; (2) delivering as high
a playback bitrate as possible within the throughput con-
straints; (3) minimizing startup delay so that the user does
not quit while waiting for the video to load; and (4) keeping
the playback as smooth as possible by avoiding frequent or
large bitrate jumps [24, 35].
To see why these objectives are conflicting, let us con-
sider two extreme solutions. A trivial solution to minimize
rebuffering and the startup delay would be to always pick
the lowest bitrate, but it conflicts with the goal of deliver-
ing high bitrate. Conversely, picking the highest available
bitrate may lead to many rebuffering events. Similarly, the
goal of maintaining a smooth playback may also conflict if
the optimal choice to simultaneously minimize rebuffering
and maximizing average bitrate is to switch bitrates for ev-
ery chunk.
The focus of this paper is on client-side adaptation solu-
tions. Other complementary work includes the use of server-
side bitrate switching (e.g., [37, 18]), TCP changes to avoid
bursts (e.g., [27]), and in-network throughput management
and caching (e.g., [31, 45, 42]). We focus on the client-side
problem for two key reasons. First, client-side solutions of-
fer the most immediately deployable alternative in contrast
to solutions that require in-network support (e.g., [31, 45,
42]), server-side software changes (e.g., [37, 18]), or modi-
fications to lower-layer transport protocols (e.g., [27, 28, 39,
29, 36]). Second, the client is often in the best position to
quickly detect performance issues and respond to dynam-
ics. That said, we believe that the formal foundations and
algorithms we develop can be equally applied to these other
deployment scenarios.
Many measurement studies have shown the poor perfor-
mance of state-of-art video players with respect to these QoE
measures (e.g., [46, 34, 32]). These studies show that most
problems are not artifacts of specific players but manifest
across all state-of-art players such as SmoothStreaming [13],
Netflix [11], Adobe OSMF [3], and Akamai HD [4]. For
brevity we do not reproduce these results here but refer in-
terested readers to prior work (e.g., [46, 34, 32]).
To alleviate these problems, there have been several recent
proposals in the research literature (e.g., [47, 34, 18, 33, 38]).
At a high level, these solutions can be roughly divided into
two categories: (1) rate-based algorithms and (2) buffer-
based algorithms. Video players with rate-based methods
essentially pick the highest possible bitrate based on the es-
timated available throughput. However, as shown in prior
work throughput estimation on top of HTTP suffers from
significant biases [32], which leads to problems with tradi-
tional rate-based approaches. Some solutions try to work
around these biases by either smoothing out throughput esti-
mates [47] or choosing better scheduling strategies [34]. On
the other hand, recent work makes a case for buffer-based al-
gorithms [33]. Rather than using throughput estimates, this
class of algorithms uses buffer occupancy as the feedback
signal, and designs mechanisms to keep the buffer occu-
pancy at a desired level, essentially discarding any available
throughput information.
Despite the broad interest in this topic, what is critically
lacking today is a principled understanding of bitrate adapta-
tion algorithms. Each aforementioned solution offers point
heuristics that work under specific (and implicit) environ-
mental assumptions. While each approach seen in isolation
has been shown to outperform commercial players, there is
little effort to systematically compare how different classes
of algorithms stack up against each other or which of these
technical components are critical, or how robust these algo-
rithms are across different operating regimes (e.g., through-
put stability, buffer size, number of bitrate levels). Further-
more, many of these algorithms even fail to formally state
what objective they seek to optimize making it harder to con-
duct a meaningful comparison.
Our first-order goal in this work is to bring some clarity
to this space. Rather than design yet another point solution,
we start by developing a first-principles approach via con-
trol theory to develop a general framework to reason about
classes of algorithms. In the next section, we use this control-
theoretic “lens” to formally define the stochastic optimiza-
tion that video bitrate adaptation algorithms try to solve.
3 Control-Theoretic Model
In this section, we develop a mathematical model of the
HTTP video streaming process and formally define the bi-
trate adaptation problem. This model gives us a framework
to compare and evaluate existing algorithms and serves as
the foundation for potential improvements.
3.1 Video Streaming Model
We model a video as a set of consecutive video segments or
chunks, V = {1, 2, · · · , K}, each of which contains L sec-
onds of video. Each chunk is encoded at different bitrates.
Let R be the set of all available bitrate levels. The video
player can choose to download chunk k at bitrate R
k
∈ R.
Let d
k
(R
k
) be the size of chunk k encoded at bitrate R
k
. In
constant bitrate (CBR) case, d
k
(R
k
) = L × R
k
, while in
variable bitrate (VBR) case the d
k
∼ R
k
relationship can
differ across chunks.
The higher bitrate is selected, the higher video quality
is perceived by the user. Let q(·) : R → R
+
be a non-
decreasing function which maps selected bitrate R
k
to video
quality perceived by user q(R
k
). Note that q(·) may depend
on the video-playing device as well as the content of the
video. For example, while on HDTV 3Mbps and 1Mbps
may lead to significant difference in user experience, the
327
剩余13页未读，继续阅读
hongye_main
- 粉丝: 11
- 资源: 3
我的内容管理
展开
最新资源
- C++标准程序库：权威指南
- Java解惑：奇数判断误区与改进方法
- C++编程必读：20种设计模式详解与实战
- LM3S8962微控制器数据手册
- 51单片机C语言实战教程：从入门到精通
- Spring3.0权威指南：JavaEE6实战
- Win32多线程程序设计详解
- Lucene2.9.1开发全攻略：从环境配置到索引创建
- 内存虚拟硬盘技术：提升电脑速度的秘密武器
- Java操作数据库：保存与显示图片到数据库及页面
- ISO14001：2004环境管理体系要求详解
- ShopExV4.8二次开发详解
- 企业形象与产品推广一站式网站建设技术方案揭秘
- Shopex二次开发：触发器与控制器重定向技术详解
- FPGA开发实战指南：创新设计与进阶技巧
- ShopExV4.8二次开发入门：解决升级问题与功能扩展
资源上传下载、课程学习等过程中有任何疑问或建议，欢迎提出宝贵意见哦~我们会及时处理！
点击此处反馈
