We present steps toward big data application design over distributed systems. We propose a HDFS block placement algorithm for the network reliability problem identified in geographically distributed topologies. The heuristic algorithm we propose provides better performance compared to the default block placement method. We simulate our solution to show improved quality.