Load-balancing for WAN warehouses

Although the basic Data Warehouse schema concept is centralized, there are increasingly application domains in which there is the need to have several sites or computers input and analyze the data, therefore distributed data placement and processing is necessary. Given that sites may have different amounts of data and different processing capacities, how can we conform to the placement requirements of the context and balance such a system effectively? In WAN environments the network speed is a very relevant factor and there are application requirements concerning the place where each piece of data stays, based on who produced the data (ownership). We propose a new strategy that accepts the placement requirements of the desired context and uses an effective automatic approach to determine fixed-sized chunks and to balance and process those chunks efficiently. Our experimental results show the validity of the approach and how to minimize the context limitations.