Graph centrality algorithms for resource allocation in distributed query processing

In this study, the effectiveness of the topological properties of the nodes has been investigated with help of the centrality based algorithms. Firstly, recently distributed centrality algorithms have been analyzed, clustered and investigated their applicability of algorithms to complex networks. Then, three centrality based resource allocation algorithms have been proposed. These algorithms are designed taking into account the needs of the distributed query processing. Proposed algorithms have some heuristic approaches to be fast and to find correct candidates. Besides, proposed algorithms are good example of their class. Finally, theoretical and experimental analyzes of the compare and proposed algorithms have been carried out. Real-life conditions are taken into consideration in simulation studies. As a result of the simulation studies, the CCBC and BCBC algorithms are very close to each other and gave the best results. This shows that closeness and betweenness centrality approaches have a positive and similar effect on the resource allocation process. Based on this; in distributed query processing systems, it has been concluded that the topological properties of the nodes are very effective and it is appropriate to use them as a parameter in the cost model used in the resource allocation phase. Besides, in this study, proposed limited centrality approach and subgraph detection approaches have been observed to affect the algorithm costs positively in the candidate determination process. These approaches, which are designed by considering the distributed query processing system and centrality calculation methods, not only narrow the candidate cluster but also contribute to the selection of the correct candidate.