The invention is suitable for any kind of distributed networks having different data rates on different connections and a centralized computer system wherein the data to distribute is created. The central system comprises one or more supersets of data that is to be distributed to client computers by extracting subsets of data from the superset/s of data. Thereafter the subsets of data are distributed to selected groups of clients based on information about the network architecture. In this way an efficient way of distributing data is executed and no duplicated data is distributed. Preferably the subsets are distributed to clients based on a counter performance or a data rate of a client's connection to the central computer system.