Efficient declustering techniques for temporal access structures

We address the problem of declustering temporal data and access structures for parallel architectures consisting of a single processor and multiple disks. To illustrate our techniques we choose the multi-version access structure. We propose two new, efficient techniques called T-proximity and KT-proximity for assigning the data and index nodes of the multiversion access structure to multiple disks. The KT-proximity declustering technique considers both the key and temporal dimensions of the data in the nodes to achieve uniform load distribution and decrease the response time for key-range, time-range and combined range queries. The T-proximity technique considers only the temporal dimension. Extensive simulations of the T-proximity and KT-proximity techniques validate their efficiency. Our results demonstrate that this technique outperforms previous methods based on random, multilevel round-robin, the LoT scheme, and proximity based on time only. Finally, we emphasize that KT-proximity is a general technique applicable to any temporal access structure