Abstract:
|
In data management for large-scale applications such as Peer-to-Peer networks, and Grid and Cloud Computing arise challenges in regard to the decentralization of the application and in regard to an increasing number of failures. A consequence of these conditions is an increasing retrieval time, inaccurate results and higher network consumption. A solution to restrain an increasing retrieval time and an increasing number of messages is the introduction of approximate queries. The introduction of approximate queries limits the querying of all nodes of a network to a subset of nodes in the cost of the results' accuracy. Thus, a conflict to provide a large-scalability lies in guaranteeing accurate data, in the provision of fast results and in a low consumption of network bandwidth. Therefore, we propose an information aggregation that is based on an analytic hierarchical process (AHP) to find a trade-off among the unpredictable factors of time, messages and accuracy. After a user defines the preferences for the retrieval process, the AHP-based algorithm makes autonomous decisions on each node. The algorithm decides autonomously about pruning the approximate queries to reach an optimal trade-off from a global perspective. Applying the self-regulated pruning of the approximate queries allows reducing the messages from an exponential in crease to a constant factor. At the same time, the retrieval time is reduced from a linear increase to a constant factor in regard to an increasing number of nodes. At the same time of reducing the retrieval time and network bandwidth, the AHP-based self-regulation guarantees certain level of accuracy |