Cluster research based on remote server contention states using K-Means over the internet

In the environment of data integration over the Internet, the remote serverpsilas contention states take direct effect on the cost of a data query. So to determine the server contention states plays an import role to estimate the cost of query. This paper uses sample queries and k-means algorithm to determine the remote serverpsilas contention states, and get the response cost of the server, then develops a set of cost model for each server contention states by a multiple regression process, to estimate the cost in the system. This method can accurately predict the system contention state and estimate the cost of a query precisely, with the acceptable error which the maximum is 26 percent, the minimum is 7.4 percent, most are around 10 percent.