Engineering Tree Labeling Schemes: A Case Study on Least Common Ancestors

We address the problem of labeling the nodes of a tree such that one can determine the identifier of the least common ancestor of any two nodes by looking only at their labels. This problem has application in routing and in distributed computing in peer-to-peer networks. A labeling scheme using Θ(log n)-bit labels has been previously presented by Peleg. By engineering this scheme, we obtain a variety of data structures with the same asymptotic performances. We conduct a thorough experimental evaluation of all these data structures. Our results clearly show which variants achieve the best performances in terms of space usage, construction time, and query time.