Routing in complex networks based on hidden metric spaces

OData support
Dr. Gulyás András
Department of Telecommunications and Media Informatics

The P2P networks play a key role at the internet, as according to several measurements, such applications have a significant share of the total bandwidth consumption.

They all build a virtual network over the internet and exploit its smart structure for routing and efficient searching.

But the frequent leaving and joining of nodes means a big challenge in the network and results in poor performance. The measurements of real-world systems show that the behaviour of users is exactly like that. However the probability of leaving for a specific node is not the same for all users within a network, so the users are not homogeneous inside a P2P network.

It's important to avoid the side-effect of the churn phenomenon and exploiting the heterogenity offers a way to minimize the performance loss caused by churn. This is the design principle of the network primitives in the thesis.

The participants of the network try to fill a graph structure which comes from tessellation in the hyperbolic space. At routing, this graph is ideal as greedy forwarding is doable without more considerations.

The basis of the search: go closer and closer to the destination with each step using geo-routing, but in the hyperbolic space.

I check the correctness and the performance of the new design by computer simulation.


Please sign in to download the files of this thesis.