Collaborative Caching in Content-Oriented Networks

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering


PhD Thesis Defence


Title: "Collaborative Caching in Content-Oriented Networks"

By

Mr. Jie DAI


Abstract

The content-oriented network is becoming a reality with enormous amount of 
contents such as high-definition videos and software packages being spreading 
across the entire network as the daily routine. Such an explosive demand on the 
content delivery has recently gained much attention with the increasing user 
popularity and the successful deployment of commercial systems. However, this 
also poses significant challenges on the existing network infrastructure due to 
the tremendous consumption of resources such as storage, link bandwidth and the 
involved network traffic cost. With the large amount of contents available in 
the entire network, how to efficient utilize the network capacity is becoming a 
critical problem in both research and practice.

The deployment of cache servers in content-oriented networks can help to 
alleviate the resource constrains, and also helps to move the content closer to 
end users with better user experiences. Cached contents are usually located 
close to users, which can greatly reduce the traffic on the network backbone 
with the improved performance. It also offloads the involved traffic cost by 
limiting the unregulated inter-domain traffic. Despite a large amount of 
existing caching mechanisms for conventional web applications, relatively 
little has been done in the cache design for content-oriented networks, which 
exhibit unique traffic features and user behaviors. Furthermore, there is a 
great potential as well as challenge in exploring collaboration among cache 
servers given their autonomous property and distributed nature.

In this thesis, we address the design and the analysis of collaborative caching 
mechanisms in several scenarios within content-oriented networks. We first 
discuss caching strategies in structured content-oriented networks. Based on 
the hierarchical topology derived from real-world Internet Protocol television 
(IPTV) systems, we propose efficient mechanisms to explore the capacity of the 
existing system infrastructure. We then investigate collaborative caching 
mechanisms in peer-to-peer (P2P) applications, considered as typical 
unstructured content-oriented networks. We specifically focus on eliminating 
the tremendous inter-domain traffic cost, with respect to dynamic P2P traffic 
patterns, peering policies and cache server capacity constraints. We then 
design incentive mechanisms for the cache collaboration in wireless multimedia 
networks, in order to promote the potential collaboration among selfish cache 
servers belonging to different administrative domains.

Through studies in both structured and unstructured content-oriented networks, 
we observe that the overall system performance can greatly benefit from the 
proposed collaborative caching mechanisms while the involved traffic cost is 
also minimized. Caching decisions are made based on specific topological 
properties, capacity constraints and optimization objectives. The analysis on 
incentive mechanisms further improves the practicability of proposed 
mechanisms.


Date:			Monday, 18 June 2012

Time:			3:00pm – 5:00pm

Venue:			Room 3501
 			Lifts 25/26

Chairman:		Prof. Wenxiong Wang (LIFS)

Committee Members:	Prof. Bo Li (Supervisor)
 			Prof. Lei Chen
 			Prof. Lin Gu
 			Prof. Chin-Tau Lea (ECE)
                      	Prof. Xiaowen Chu (Comp. Sci., Baptist U.)


**** ALL are Welcome ****