Family trees: an ordered dictionary with optimal congestion, locality, degree, and search time

We consider the problem of storing an ordered dictionary data structure over a distributed set of nodes. In contrast to traditional sequential data structures, distributed data structures should ideally have low congestion. We present a novel randomized data structure, called a family tree, to solve this problem. A family tree has optimal expected congestion, uses only a constant amount of state per node, and supports searches and node insertion/deletion in expected O(log n) time on a system with n nodes. Furthermore, a family tree supports keys from any ordered domain. Because the keys are not hashed, searches have good locality in the sense that intermediate nodes on the search path have keys that are not far outside of the range between the source and destination.

• Type: Conference
• Year: 2004
• Where: SODA
• Authors: Kevin C. Zatloukal, Nicholas J. A. Harvey
• Keywords: Algorithms, Data Structure, Dictionary Data Structure, Family Tree, SODA 2004