In a data-indexed DHT overlay network, published data annotations form distributed databases. Queries are distributed to these databases in a nonuniform way. Constructing content distribution networks for popular databases is effective for addressing the skew problem. However, various crucial replication decisions such as which data object should be replicated, how many replicas should be created and what replica placement policy should be used, may affect the performance of replication based content distribution and load balancing mechanisms in the presence of non-uniform data and access distribution. Particularly, the impact of the propagation speed of replicas on the performance of such type of overlay networks is not well studied. In this paper, a proactive method is given to tackle this problem. The method can adaptively adjust the number of replicas to create based on the changing demand. It works in a fully distributed manner. Our experiments show that the approach is able to ad...
Bassam A. Alqaralleh, Chen Wang, Bing Bing Zhou, A