| The thesis proposes the use of learning algorithms to achieve self-organization network intelligence. We focus our analysis on bulk file transfer optimization. Our proposed Bulk File Transfer procedure is leveraged to optimize the throughput in communication networks. For executing real-time changes in our network topology, we used the principle of activity-dependent topology changes and correlation in topology structure. The paradigms we consider here are responsibility distribution among different nodes, their implicit coordination, trust of nodes, and traffic control in a network. We denote our intelligent network protocol as Cognitive Self-Organization (CSO). CSO requires a prior framework for Bulk File Transfer which is obtained using the Concurrent File Transfer Protocol. To facilitate network self-organization, we propose a unique address assignment scheme for every node. We particularly focus upon wireless control plane frameworks for cloud data center networks. The network throughput obtained for every job in the Bulk File Transfer framework is compared with the one obtained in the Self-Organized Bulk File Transfer framework. Our results show that, the network throughput for most of the jobs in the network is improved when the Self-Organized framework is considered. |
