University of Surrey

Test tubes in the lab Research in the ATI Dance Research

Improving Performance of P2P Networks Through Semantic Topological Adaption.

Eftychiou, Athena. (2013) Improving Performance of P2P Networks Through Semantic Topological Adaption. Doctoral thesis, University of Surrey (United Kingdom)..

Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (10MB) | Preview


In today's world a tremendous amount of information is produced daily. The plethora of data represents a challenge in terms of how to manage, represent and share it, in a meaningful and effective way. This increasing size of on-line information and its weak structuring, suggests the need for principles to make information accessible through well-defined operations on well-defined data. The mixture of P2P and semantic technologies could harvest the advantages of both mechanisms and address the challenges of on-line information management and sharing. P2P networks combined with semantic technologies promise to achieve distributed management and exchange of information in an efficient manner. On the one hand, P2P computing offers a more effective alternative to existing client-server applications by providing a radical way of decentralised information management and sharing. On the other hand, semantic-oriented technologies present new approaches in solving the problems of information complexity by adding meaning and structure to data, in order to facilitate better information management, indexing and retrieval. This thesis proposes a system for distributed information management and sharing in unstructured P2P networks through the utilisation of semantic information, extracted from the network resources. Its aim is to facilitate an efficient way of information exchange while keeping low messaging cost and normal load balancing among the peers of the network. In particular, the proposed architecture follows a two-layer approach where the upper layer forms the semantic knowledge of the network through super-peers, and the lower layer of peers represents the network resources. The network knowledge is formally represented by a domain specific ontology using collective intelligence techniques to extract knowledge from the available resources. This semantic-driven approach along with dynamic topological adaptations is facilitated by two key components: the Semantic-Driven Architecture (SDA) and Dynamic Adaptive Topology (DAT). SDA aims to improve query efficiency and achieves this by mapping the network ontology to the overlay topology creating in this way a semantic-driven architecture. During the resource discovery process the query is intelligently routed in the semantic layer via ontology supported decisions and attaining in this way better query success and reduced network traffic. DAT introduces dynamic topological optimisation on top of the SDA component having as objectives scalability and fault tolerance. Since DAT component is based on SDA model, topological adaptations follow a semantic-driven approach for retaining the successful architecture of SDA. Scalability is achieved through load balancing mechanisms, where overloaded or under-loaded super-peers are optimised accordingly, and the overall semantic image of the network is retained up to date, through reconceptualisation procedures. A number of experiments were carried out and showed that the proposed model demonstrates high success rate and low traffic, and therefore outperforms two existing popular P2P paradigms used as benchmarks, Social P2P and Traditional super-peer architecture. In particular, SDA and DAT demonstrate a success rate in the range of 80-90% as opposed to Traditional and Social super-peer based P2P which achieve 65% and 55% success rate correspondingly. The experimental results have confirmed both the soundness of the design of the SDA model, since the ontology-topology mapping with combination to query-ontology mapping, consequently lead to the successful query-topology mapping; and have also verified that a dynamic adaptive model can achieve load balancing and scalability through semantic topological optimisations.

Item Type: Thesis (Doctoral)
Divisions : Theses
Authors : Eftychiou, Athena.
Date : 2013
Additional Information : Thesis (Ph.D.)--University of Surrey (United Kingdom), 2013.
Depositing User : EPrints Services
Date Deposited : 24 Apr 2020 15:26
Last Modified : 24 Apr 2020 15:26

Actions (login required)

View Item View Item


Downloads per month over past year

Information about this web site

© The University of Surrey, Guildford, Surrey, GU2 7XH, United Kingdom.
+44 (0)1483 300800