In this paper, we propose, design and simulate a class of distributed multi-sensor data-fusion systems, which consist of one data fusion center (DFC) and many local sensors that are connected via local area networks (LAN) and/or wide area networks (WAN). Each local sensor has its own decision maker. These local sensors may use various sensing techniques, e.g. radar RF sensors, IR sensors, and sonar sensors. These systems can serve ocean surveillance for the purpose of processing in real-time large volume of data which the sensors receive from the environment, e.g. multiple sonars are used to detect submarines. Employment of multi-sensor for surveillance has the advantages of reliability, survivability, and capacity of coverage. The system is real-time in the sense of application layer. The critical time constraints are applied in local processors, transferred data packets and DFC. We developed a sophisticated communication interface and protocols such that this system can get high throughput, short response time and high reliability.