关键词: 聚类, 围绕中心点的划分（PAM）算法, 并行, 消息传递接口（MPI）

Abstract: Parallel is very important in data mining.This paper proposes a distributed k-mediods clustering algorithm by analyzing how to get satisfactory clustering information from local information.Its quality is equivalent to serial PAM algorithm but its calculation performance is higher.The paper gives still the way that improves the efficiency of parallel PAM algorithm by reducing the cost of communication，analyzes the performance of the algorithm and gives the result of experiment.This explains that the algorithm is effective and reliable.

Key words: clustering, Partitioning Around Mediods（PAM） algorithm, parallel, Message Passing Interface（MPI）