Distributed suffix array construction algorithms: Comparison of two algorithms

The suffix array is an important indexing data structure for biological sequence analysis. The increasing size of genomic data necessitates the use of a computer cluster to speed up the computation. In this paper, we compare the performance of two distributed suffix array construction algorithms that can run on a computer cluster: the first is Futamura-Aluru-Kurtz algorithm and the second is Kulla-Sanders's. The performance of these two algorithms has not been compared earlier due to the lack of any available software and the difficulty of implementing them. In this paper, we answer a still open question about which algorithm would be better in practice, particularly for genomic data. We have implemented the two algorithms and made them available. The comparison results show that the Futamura-Aluru-Kurtz algorithm is more efficient in practice, despite the fact that the algorithm of Kulla-Sanders has better theoretical time complexity.