At present, although methods of automatic, periodical manual data collection have been adopted in some areas, there are many problems existing in these methods. This paper proposes a new framework for online monitoring of water environment based on overlay network technology. Especially, we build the multi-level overlay network, which consists of the GPRS networks, mobile networks and internet networks. Following this, a multi-dimensional data cube for water environment is established using the ETL process with the input of complex heterogeneous data collected. Thereby, the framework of data center is established for on-line early warning of water environment and data analysis and processing. This framework has been put into use in Wuliangsuhai for the on-line real-time water monitoring tests. The results show that the overlapping network architecture is effective and on-line analysis and early warning of heterogeneous data is efficient.