An efficient external sorting algorithm

Abstract

This paper presents an algorithm for external sorting with two-level memories. We only consider the number of disk I/O because the disk I/O is the bottleneck of external sorting. OUr method is different form the traditional merge sort and uses the sampling information to reduce the disk I/O in the external phase. Our algorithm is elegant, simple and making good use of memory available in the computer environment now. Under certain memory constraint, the algorithm runs with optimal number of disk I/Os, where N is the number of records to sort and B is the block size