Implementing Distributed Search Structures

Distributed search structures are useful for parallel databases and in maintaining distributed storage systems. In this paper we discuss some issues in the design and implementation of distributed B-trees, such as methods for low-overhead synchronization of tree restructuring, node mobility and replication strategies. We have implemented a distributed B-tree that replicates the index and also performs data load balancing. Replication of the index provides for high throughput access, and data balancing allows for balanced processor and space utilization. For replication, we propose two diierent strategies, namely, full replication wherein the entire index is replicated at each processor and path replication where if a processor owns a leaf node, it also holds the index nodes on the path from the root to that leaf node. We present an algorithm for dynamic data-load balancing which uses node mobility mechanisms. We present some performance results of our algorithms. We nd that path replication will create a scalable distributed B-tree.