Maximal Profit Strategy Of Multi-agent Scheduling System Meeting User’s Security Requirements

With the rapid development of large-scale heterogeneous network and traditional network management framework, scheduling technology and optimization algorithm have been unable to meet the needs of the development of network. Especially under the background of network integration promoted by the technology represented in cloud computing today, large-scale computing network distribution, heterogeneous resources, and other significant characteristics are formed, which makes it difficult to apply the single resource scheduling management system and related algorithms. In addition, with the rapid development of Internet applications, the feature of diverse performance requirements is increasingly outstanding. The user’s requirements for the timeliness, reliability, data security of the service are higher and higher, however in order to improve the enterprise operating income, the service provider should improve resource utilization, save energy to reduce the cost on the basis of meeting the requirements of users. These problems are currently urgent to solve.Multi-agent system theory is applied in this thesis, the multi-agent scheduling system model is established in order to break the limitations of single resource scheduling management system. The model forms a new network management framework of easy extension, efficient management, dynamic scheduling. The framework will be applied to the large-scale network environment, which is distributed heterogeneous and dynamic extension. Multi-agent scheduling system model is mainly composed of global agent and execution agent. Global agent is responsible for receiving the user’s task information, execution agent is responsible for executing task. Considering the profit of service provider, global agent will choose the execution agent which has the maximal profit to perform the task. Considering the user’s requirements for the service reliability and data security, security level is made in this thesis by adding protection resources to meet the needs of user. Towards the execution agent which can’t meet the demand of security level, the system will add protection resources into execution agent to ensure normal operation of the task. Towards execution agent which meets the demand of security, protection resources are not needed. Because of the fault of resources, the service time is random. Universal generating function is very effective to evaluate the multi-state variable, so universal generating function is adopted in this thesis to calculate the service time, combining with the service price list, the service’s price can be determined easily. Service profit is mainly influenced by task partition and resource allocation. Only the optimal method of task partition and resource allocation is found, the largest service profit can be obtained. Genetic algorithm is very suitable for solving such optimization problem, so genetic algorithm is used in this thesis to solve the profit maximization problem. At the end of this thesis, experiments are operated in two cases that ignoring security level and meeting the security level to obtain the largest service profit, meanwhile the experiments under different security levels are also operated, and the experiment results are compared and analyzed. The experiment results show that the profit maximization strategy proposed in this thesis is feasible and effective.