Research On The Performance And Reliability Of Lightweight Publish/Subscribe System Based On IoT

In the Internet of Things environment,devices often have low processing capabilities,network bandwidth,and stability.Some devices use battery power,which also imposes certain requirements on the energy consumption of the system.These make the Internet of Things have higher performance and reliability requirements for message transmission.The publish/subscribe model becomes a communication model of the Internet of Things because of its good loose coupling.MQTT has become a widely used publish/subscribe protocol due to its simplicity,low overhead,and high reliability.This thesis firstly analyzes the MQTT protocol in detail,and puts forward an improved scheme to address the shortcomings.Then,based on the characteristics of the publish/subscribe system,it proposes an agent distribution architecture.Finally,it measures the agent load and proposes an agent assignment algorithm.The main works of this thesis are as follows:1)Analyze the shortage of the MQTT protocol retained message storage distribution mechanism,and propose an improved retain message processing mechanism based on the subscriber's problem of missing retain messages and receiving retain messages repeatedly.Optimize the retain message storage structure,modify some packets of the protocol and the corresponding processing mechanisms of the server and the client so that the subscriber can receive all retain messages that are missed during the disconnection after the subscription,and do not repeat receiving the latest retain message.After that,the improved retain message processing mechanism is tested and analyzed.The results show that the improved mechanism solves the problem of missing retain messages and repeatedly receiving retain message.2)Analyze the publish/subscribe system architecture model and propose a publish/subscribe architecture model based on centralized distribution.The model dynamically assigns a proxy server to the client,and centrally distributes the publisher or subscriber under the same topic to the same or a nearby proxy server,minimizing the consumption of messages transmission between the agents.At thesame time,the solutions to node overload and node failure are given to ensure the stability of the system.After that,comparing the performance of this model with a randomly selected distribution agent scheme through experimental tests.The results show that in the same situation,adopting the model for agent distribution has a lower message transmission delay.3)Analyze the characteristics of the publish/subscribe system and propose a dynamic agent distribution algorithm based on historical load.The factors affecting the agent load are analyzed and quantified to obtain the agent load evaluation model.Combined with the characteristics of the publish/subscribe system,the influence of the historical load on the agent node is considered,making the distribution of the agent more explicit and effective.After that,through experimental tests,compares the performance of the system under random distribution,distribution based on load,and distribution based on historical load.The results show that the dynamic agent distribution algorithm based on historical load can better reduce the message transmission delay and improve overall system performance.4)Applied the above improved schemes to the actual Internet of Things project to test their actual availability.The results show that the above improved schemes are actual feasible.