Semi-fluid: a content distribution model for faster dissemination of data
Loading...
Date
2010
Authors
Saleh, Salah Noori
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This thesis proposes and implements a novel content distribution model for
reducing or minimizing delay in data dissemination. Currently, content distribution is
based on two models: the Fluid model and the Chunk model. The Fluid model
provides continuous transferring of the content from the source to multiple receivers:
For high throughput, a receiving node should distribute a bit once it has received that
bit. However, working with the Fluid model in a heterogeneous network needs
special care because the model incorporates tightly coupled connections between
adjacent nodes. This imposes fundamental performance limitations, such as dragging
down all transfer rates in the system to the rate of the slowest receiving node. In the
Chunk model, contents are first chopped into pieces of equal size and the subsequent
distribution happens in pieces. That is, a node will not distribute a piece until it has
fully received that piece. A Chunk model is a loosely coupled connections; a node
will not distribute a chunk until it has fully received that chunk, making nodes wait
to receive the entire chunk before they can start distributing it. This becomes
untenable because content transfer may take a long time and during this time the
upload capacity of downloading nodes is unutilized. Delay is critical for real-time
and interactive applications. A poor content distribution model could result in
considerably longer distribution time, while a good model could shorten the
completion time and efficiently utilize resources like network bandwidth. The novel
Semi-Fluid content distribution model proposed in this thesis will distribute chunk
content in different heterogeneous networks in a fluid manner, without having any
backpressure caused by Fluid content distribution model, or encountering chunk
transition delay caused by Chunk content distribution model, by optimizing the
existing (Chunk and Fluid) content distribution models, and enabling better
utilization of node's resource, such as local storage and bandwidth. Mathematical
proof and real implementation test results show that our proposed Semi-Fluid content
distribution model finds an optimal solution for all cases tested in heterogeneous
networks.