Research And Design Of Message Transmission Mechanism In Cloud Environment

With further development of the Internet,the scale of the cloud computing system is expanding,whoes business becoming increasingly complex.Different business forms different subsystems.As a carrier,message system is widely used in loosely coupling and asynchronous message interaction between subsystems.if using the traditional message transmission mechanism to deal with complex business scenario,may produce some unexpected errors,and bring down the entire business system.Therefore,as an important part of cloud computing technology,the research and development of message transmission mechanism has an important realistic significance.Firstly,we research the scalability and high availability of message transmission mechanism.We ensure the scalability of Broker cluster while introducing group and topic in the scalable architecture,which achieves a linear extension of the publisher and subscriber.Mean while,this paper designs a message storage backup strategy based on improved consistent hashing,which makes the message distribute more balanced,thus quickly completes fixing position and search.In order to improve the high availability,we proposes a Raft consistency protocol based on timestamp,which achieves the smooth heat switching and reduces the loss of messages when master-standby switching.Then,according to the above objectives,a series of design and optimization schemes are put forward to solve the problem about the performance of message transmission and storage.Because of low efficiency in the traditional database storage,this paper presents an alternative scheme of file storage based on Bitcask model,which is more efficient than random read/write database storage.The global route management of message transmission is designed,it not only realizes the loose coupling,but also improves the speed of message addressing.Next,we propose a finer grained double message load balancing mechanism and a message push-pull model based on persistent connection.The different load balancing algorithm based on multi queue is adopted both in publisher and subscriber,which effectively disperses client connection pressure and message transfer pressure to each Broker node.The long connection model guarantees the real time of the message transmission,and reduces the pressure of the message server.Finally,simulating and implementing the system,and testing it in a PaaS cloud computing platform,the results show that the system has a significant improvement on related performance indicators.