Multi-objetive evolutionary algorithms for unstructured P2P topologies reconfiguration
Resumen: Nowadays one of the main usages of the Internet is content generation, sharing and access. One of the approaches used to support these activities is the P2P model. In content distribution systems using this model, instead of having servers to distribute the contents, nodes are self organized to do it. Each node is a potential server and a client and this characteristic makes those systems naturally scalable and fault tolerant. In order to coordinate their actions, the nodes of a P2P system establish logical connections between them and those connections define an overlay network (called P2P network). There are two main classes of P2P networks: structured and unstructured. We are interesting on the second class, where there are no specific rules to define the neighborhood of a newcomer.
The reconfiguration of a P2P network is generally performed through a protocol to integrate new peers and a protocol to deal with those that leave the system. The goal of these protocols is to preserve the connectivity. A reconfiguration can also address the change in the interests of users, by example, by adjusting the amount of links of a peer to the popularity of its shared resources or by allowing a peer to change its neighborhood when this one is not interesting for it anymore. The reconfiguration of an unstructured P2P network can be seen as the problem of selecting, among all the possible connected undirected labeled graphs, the optimal one for the current participant peers and their interests. The prohibitive computational costs of an exhaustive approach provide an opportunity to use evolutionary computation to adaptively optimize the topology. In this project we are interested in the exploration of distributed control, local information multi-objective evolutionary algorithms as a strategy to perform periodic reconfiguration of unstructured P2P topologies.
- The design and implementation of an effective an efficient multi-objetive evolutionary algorithm to perform the reconfiguration of an unstructured P2P topology.
- Identify and understand the different distributed control, local information multi-objetive evolutionary algorithms proposed to perform the reconfiguration of an unstructured P2P topology.
- Design in a distributed control, local information multi-objetive evolutionary algorithm to perform reconfiguration in unstructured P2P topologies.
- Evaluate the proposed algorithm.
- Assess the relative importance of the different objetives used in the reconfiguration of an unstructured topology.
Última actualización: June 2, 2016 at 15:51 pm