Practical Suffix Tree Construction

Large string datasets are common in a number of emerging text and biological database applications. Common queries over such datasets include both exact and approximate string matches. These queries can be evaluated very efficiently by using a suffix tree index on the string dataset. Although suffix trees can be constructed quickly in memory for small input datasets, constructing persistent trees for large datasets has been challenging. In this paper, we explore suffix tree construction algorithms over a wide spectrum of data sources and sizes. First, we show that on modern processors, a cache-efficient algorithm with O(n) complexity outperforms the popular O(n) Ukkonen algorithm, even for in-memory construction. For larger datasets, the disk I/O requirement quickly becomes the bottleneck in each algorithm’s performance. To address this problem, we present a buffer management strategy for the O(n) algorithm, creating a new disk-based construction algorithm that scales to sizes much larger than have been previously described in the literature. Our approach far outperforms the best known diskbased construction algorithms.