Data mining from relations is becoming increasingly important with the advent of parallel database systems. In this paper, we propose a new algorithm for mining association rules from relations. The new algorithm is an enhanced version of the SETM algorithm (Houtsma & Swami 1995), and it reduces the number of candidate itemsets considerably. We implemented and evaluated the new algorithm on a parallel NCR Teradata database system. The new algorithm is much faster than the SETM algorithm, and its performance is quite scalable.