| Summary: | 碩士 === 國防大學理工學院 === 資訊工程碩士班 === 102 === Nowadays, the major high level command and control systems still use the client-server network architecture to perform on tactical network. This traditional architecture usually causes client and server-side network bottleneck, and server failure tends to cause of the service crashes. The central servers have always been attacked for the primary targets in the warfare. In tactical circumstances of rapid changes, peer-to-peer (P2P) networks can be adapted for seamless scalability, high-availability and taking advantage of decentralized resources.
In our study, we use a P2P distribution computing model using irrequally collected peer information, and the developed a peer hierarchical management mechanism that composes geographical location-based clustering P2P network dynamically for using under-utilized computing resources efficiently. The simulated experiment environments use JXTA protocols, and simulated observe collaborative access, resource discovery and resource allocation among P2P nodes. We use joint air and missile defense for the scene, which needs to consider quick-response features of the tactical network environment. Meridian algorithm performs central node selection based on network location, using direct measurements of average latency between two nodes. While the meridian approach is applicable for a loosely structured overlay network, it is not guaranteed that the central node is optimal performance node. In this study, we proposed an objective-function-based meridian algorithm for a supernode election, backup mechanism of supernode candidate and efficient resource assignment. We also proposed a key performance objective function and designed some experimental models compared with meridian algorithm and a random node selection.
Experimental results show that our approach can achieve supernode for better performance that assigns firepower allocation jobs in each partitioned area. Comparison of simulation results and observed data revealed that proposed supernode election algorithm could well from a radar detector, firepower assignment to notify firepower launches. Our proposed algorithm showed significant improvement the overall response time on the posttest compared to the meridian algorithm and the random node selection. The proposed algorithm has better overall response time with much smaller 1.6 to 4.3 seconds in comparison to the others. We believe that our approach provides a valid solution to the problem of dynamic resource allocation and ensuring stable operation of missile systems, further, to speed up interception efficiency of missile system.
|
|---|
