Research On Transaction Middleware Server Architecture Based On SEDA

In large scale distributed enterprise environment, applications always face with different network environment, hardware platform and communication protocol. If applications need to solve the problems caused by communications between servers, there is no doubt that the difficulty and efficiency of development will be greatly influenced. But the using of middleware provides an easy way to solve this problem.For the features of civil aviation business request such as complicate, large data amount, high concurrency and close connection, the software architecture of middleware server is designed. The merits and drawbacks of several prevalent concurrent server models are analyzed and Staged Event‐Driven Architecture(SEDA) is selected to be the base of server architecture. Meanwhile, civil aviation businesses are divided into stages according to the idea of SEDA, and the deployment and operation structure of middleware platform under distributed environment are designed. The Staged Event‐driven characteristic of middleware platform is supported by two layers of SEDA. This architecture makes application developers have no need to care about the details of asynchronous call, which greatly reduces the complexity of application development, and improves the efficiency of development. Finally, an experiment is carried out to prove that servers which have applied SEDA have better performance than those based on thread pool model when dealing with high concurrency requests, which can make Passenger Service System respond effectively to high load and concurrency environment.Meanwhile, the architecture of current system is analyzed, the fixed scale factor load balance algorithm is put forward based on it and the robustness of the system is analyzed. To promote the robustness and reliability of civil aviation business system, a dynamic load balance strategy is put forward based on the fixed scale factor load balance mechanism in accordance with the software and hardware feature. Meanwhile a new data structure and the relative method to maintain measurement indicators on it are designed. Finally, an experiment is carried out. The consequences indicate that the load balance strategy can help promote the robustness and reliability of the system.