The problem of multiple data sources selection (MDSS) in DSE (data-sharing environments) is addressed and the algorithm MDSSA (MDSS algorithm) is presented. MDSSA introduces the concept of Pareto optimization which reduces the search space greatly. By means of a novel normal-measure based nonlinear cost function, MDSSA computes approximate Pareto optimal paths to each data source first, and then gives the optimal data source and its corresponding path by comparing the cost of all candidate paths, resulting in finding more effective paths and much shorter response time. Extensive simulations show the efficiency of the algorithm.