Catálogo de publicaciones - libros

Connectivity based clustering has wide application in many networks like ad hoc networks, sensor networks and so on. But traditional research on this aspect is mainly based on graph theory, which needs global knowledge of the whole network. In this paper, we propose a intelligent approach called spreading activation models for connectivity based clustering (SAMCC) scheme that only local information is needed for clustering. The main feature of SAMCC scheme is applying the idea of spreading activation, which is an organization method for human long-term memory, to clustering and the whole network can be clustered in a decentralized automatic and parallel manner. The SAMCC scheme can be scaled to different networks and different level clustering. Experiment evaluations show the efficiency of our SAMCC scheme in clustering accuracy.

Efficient resource allocation for submitted task is one of the main challenges in grid computing. Much work in this aspect has focused on resource discovery and usage. Resource allocation, which is an interface between resource discovery and usage, is always unvalued in the past while some resource owners may provide resources with poor quality. A personalized fair reputation scheme is proposed in this paper to overcome these limitations and to encourage collaboration in grid community. We present minimum and maximum threshold for reputation from other entities to decrease malicious competition. A rewarding mechanism is also proposed in our reputation scheme to encourage collaboration. Then we present how to integrate our reputation scheme into grid resource allocation approach. Experimental results show that our approach can work in the environment with high ratio of malicious nodes in grid.