Simulation And Performance Analysis Of Distributed Publish Subscribe System

Distributed publish-subscribe system with distributed server framework has gained more and more attention for its essential contribution to distributed system. However, distributed publish-subscribe system is complicate to implement, hard to deploy and control while its performance is affected by multiple setting parameters. This makes it difficult to study and improve practical systems. Moreover, so far we still cannot find any authoritative performance evaluation method for distributed publish-subscribe system. To change such situation, in this paper we model distributed publish-subscribe system and implement it on simulation platform. We also analyze the performance of distributed publish-subscribe system and propose a performance evaluation method.Considering the representativeness, this paper focuses on distributed topic-based publish-subscribe system. The system is modeled and implemented on simulation platform NS3, and the performance of the simulating system is analyzed to retrieve the performance evaluation method for distributed publish-subscribe system. In this paper, the architecture and message process of message middleware especially publish-subscribe system are firstly discussed. Then the model of distributed topic-based publish-subscribe system is designed according to the relevant research. The final model is implemented on NS3and some running instances and data analysis of the simulation system are illustrated. Simulation results show that the simulating output data can well reflect the performance of distributed topic-based publish-subscribe system, and the provided interfaces of the system make it easy to improve system. Finally, performance of the simulation system is analyzed to retrieve performance evaluation method under different parameter combinations. We also do some testing for this evaluation method. Testing results show that the proposed performance evaluation method in this paper can precisely forecast the trend of message delay and evaluate system’s running performance.