Analytical Framework for Streaming over TCP
METADATA ONLY
Loading...
Author / Producer
Date
2010-12
Publication Type
Report
ETH Bibliography
yes
Citations
Altmetric
METADATA ONLY
Data
Rights / License
Abstract
Streaming over TCP is practical and widely used in commercial multimedia applications such as YouTube. Evaluations show commercial multimedia applications are suffering from providing good Quality of Experience [1]. Given the dynamic of Internet and the variability of TCP throughput, we propose an analytical framework for TCP streaming to trade the probability of discontinual playout against delay and buffer size. Guided by our proposed analytical framework, it is possible to allocate buffer size and initial buffer delay to provide high streaming quality for TCP streaming applications with required underflow/overflow probability and low delay, or vise versa. In the analytical framework, we propose a TCP congestion window model and TCP streaming system model, and find that the TCP window bounds1 are critical for underflow/overflow probability of TCP streaming buffer. Relying on the Gamma distribution, we derive an analytical solution for the distribution of congestion window bounds and futher an analytical solution for distribution of all window sizes. In spite of the simplicity of the model, the proposed solutions provide surprisingly accurate approximations. More importantly, the underflow probability and the overflow probability are deduced from the TCP window bounds for a given streaming buffer size and initial buffer delay. We verify the model of TCP window and the model of streaming system by extensive simulations.
Permanent link
Publication status
published
External links
Editor
Book title
Journal / series
Volume
333
Pages / Article No.
Publisher
ETH Zurich, Computer Engineering and Networks Laboratory
Event
Edition / version
Methods
Software
Geographic location
Date collected
Date created
Subject
Organisational unit
03234 - Plattner, Bernhard (emeritus) / Plattner, Bernhard (emeritus)
Notes
Funding
Related publications and datasets
Is previous version of: