Content cache and forward mechanisms in mobile networks

Get Complete Project Material File(s) Now! »

Research objectives

Heterogeneous mobile networks raise new challenges for content distribution due to the rapid  growth in number of users and the dynamics of human behaviors. These networks are large in number of hosts and highly unpredictable. This introduces more difficulty for provisionning supported infrastructures, hence there is an issue of scalability for service providers. In this context, the use of device-to-device communication can be a solution to avoid the congestion at mobile gateways. However since mobile devices have constraints in their ressources (battery life, bandwith…), these issues should also be taken into consideration.
Cooperation among users to replicate and distribute contents via device-to-device communication in such way to reduce latency and avoid congestion at gateways is highly appreciated. Therefore our objective is to design an efficient mechanism that works in this cooperative condition.
Another issue in this application context is that since there are constraints in mobile devices, it is rational to assume that users will behave selfishly, hence we should focus on the development of strategies that can be implemented in a practical network setting.

Content distribution in heterogeneous mobile networks

In heterogeneous mobile networks, content can be delivered to users either via deviceto- device communication or from a 3G connection. If a content is very popular and every users want to fetch it, a content distribution scheme using epidemic forwarding, e.g nodes  just look for content when they are in contact range, should be useful due to the following reasons :
– Every user is interested in the content, hence the availability of content is high. This can reduce the delay to download the content and the effort to look up for it.
– There is a congestion problem at the 3G service provider gateway if every user try to fetch the content from 3G.
In contrast, if only a few of users are interested in the content , there would be no congestion for users to download directly from the Internet by using 3G. The interesting problem comes when there are contents whose popularity is not high but is not as low as the congestion problem can be neglected. In this context, a replication scheme can be useful due to the following reasons :
– Replication helps increasing the availability of content. This reduces the delay in looking up and retrieving contents, hence encouraging users to switch their choice to use device-to-device communication.
– Replication helps reducing the concurrent number of downloads from the Internet, hence alleviating the congestion at 3G gateways and increasing network scalability. For the replication scheme in mobile networks, we need an efficient design to place the content replica where the content demand is. Furthermore, since the client-server model is not applicable is this case, we need a P2P mechanism to dynamically distribute the replica role to users.

P2P mechanisms

To keep the load balanced among users’ devices we need a mechanism to share the burden of content replication. This mechanism should be distributed, with low overhead and no requirement of a global view to match the unstructured nature of heterogeneous mobile networks. Human mobility may change the network topology very frequently hence the designed mechanism needs to be efficient in dealing with highly dynamic environment.
A P2P mechanism that is based only on random peer selection would be a good candidate in this context. We aim to study random content hand-over mechanisms and their performance to see if such a solution can be deployed in practice.

Optimization in content replication

Replication mechanism in mobile networks should be done in such a way that enhances content availability and reduces content retrieval latency. To do this, the problem is to find the number of replicas needed in the network and the locations to place these replicas. Given the network topology, this problem can be studied through the lenses of facility location theory. Since facility location problems are NP-hard, we need a distributed mechanism to approximate the solution in the conditions that only local information is available. Given the problems mentioned above, our work aims at finding a solution for content replication that matches the dynamic nature of mobile networks. We focus particularly on a lightweight and practical mechanism that is efficient and based only on local measurements in order to keep low overhead, while achieving good performance in terms of load balancing and content retrieval delay.

READ  Similar third-party payment systems in China

Thesis organization

The remaining of this thesis is organized as follows. In the next chapter we introduce our problem background and present a list of related works. Chapter 3 describes the problem caused by human mobility and the need to evaluate mobile application performance against realistic mobility model. In Chapter 4 we examine the mechanisms that allow users to share the burden of storing content. In Chapter 5 we study the distributed mechanisms to replicate content in mobile networks while evaluating the performance through the lens of facility location problem. In Chapter 6 we study the replication scenario when users are selfish and tend to minimize their own cost. In Chapter 7, we summarize the results of our study and outline directions for future work.

Table of contents :

List of Figures
List of Tables
1 Introduction 
1.1 Content replication in the Internet
1.2 Content replication in mobile networks
1.2.1 Mobility
1.2.2 Energy constraints and load balancing
1.2.3 Content availability
1.2.4 Selfish peers
1.3 Research objectives
1.3.1 Content distribution in heterogeneous mobile networks
1.3.2 P2P mechanisms
1.3.3 Optimization in content replication
1.4 Contributions
1.5 Thesis organization
2 Background and problem statement 
2.1 Content replication in mobile networks
2.1.1 Problem statement
2.1.2 Related works
2.2 Facility location problem
2.3 Facility location variants
2.3.1 k-median problem
2.3.2 Uncapacitated facility location
2.3.3 Capacitated facility location
2.3.4 Multiple commodity facility location
2.4 Approximation solutions for facility location problem
2.4.1 Greedy heuristic
2.4.2 LP rounding technique
2.4.3 Primal-dual technique
2.4.4 Local search technique
2.5 Replication from facility location perspective
2.5.1 Problem formulation
2.5.2 Distributed solution
3 Mobility models for wireless networks
3.1 Random mobility models
3.1.1 Random Walk
3.1.2 Random Waypoint
3.1.3 Random Direction
3.1.4 Gauss-Markov model
3.1.5 Issues in random mobility models and the Random trip model
3.2 Non-random mobility models
3.3 Human mobility
3.3.1 Mobility trace studies
3.3.2 The heavy-tail in inter-contact time distribution
3.3.3 The power-law of human mobility in virtual world traces
3.3.4 Levy flights similarity
3.3.5 The SLAW mobility
3.4 Conclusion
3.5 Relevant publication
4 Content cache and forward mechanisms in mobile networks
4.1 System objectives
4.2 Related work
4.3 P2P cache and forward mechanisms
4.4 Experimental set-up and methodology
4.4.1 Nodes placement
4.4.2 Node mobility
4.4.3 Parameter space
4.4.4 Evaluation metrics
4.5 Simulation results
4.5.1 Spatial distribution of content
4.5.2 Load balancing
4.5.3 Information access distance
4.6 Contribution
4.7 Relevant publication
5 Distributed solution for content replication in mobile networks
5.1 Problem formulation
5.1.1 Single commodity
5.1.2 Multi commodity
5.1.3 Content popularity
5.1.4 Discussion
5.2 Cost definition
5.2.1 Opening cost
5.2.2 Service cost
5.3 Distributed mechanism for replication and placement problems
5.3.1 Replica placement
5.3.2 Content replication
5.4 Simulation set-up
5.5 Single content
5.5.1 Replication with single content
5.5.2 Load balancing
5.5.3 Convergence time
5.5.4 Adaptation to demand change
5.6 Multiple contents
5.6.1 Replication with multiple contents
5.6.2 Impact of mobility
5.6.3 Scalability
5.6.4 Replica allocation
5.7 Content access mechanisms
5.8 Performance vs. the epidemic content distribution
5.9 Conclusion
5.10 Relevant publication
6 Content replication in selfish environment
6.1 Problem modeling
6.2 Socially optimal cost
6.3 A two player game
6.4 The n-player game
6.5 Contribution
6.6 Relevant publication


Related Posts