Research And Implementation Of Runtime Environment In Distributed Transaction Processing Monitor

Transaction processing technology is a key technology to ensure reliability and consistency of information. The transaction processing monitor is a kind of middleware that specially provides this technology. It offers a complete platform for developing, deploying, running, and managing three-tier based online transaction processing applications, and plays a crucial role in distributed applications.X/Open DTP Reference Model is a de facto industrial standard in distributed transaction processing, which provides the software architecture for distributed transaction applications. It is a key issue to provide a good runtime environment for distributed transaction applications conforming to this model. This thesis introduces the design and implementation of runtime environment in distributed transaction processing monitors according to X/Open DTP Reference Model, using our experience in developing the distributed transaction processing monitor OnceTX as a case study. OnceTX runtime environment not only implements an interface which conforms to the XATMI specification, but also extends it to provide more kinds of communication means in distributed transaction, such as event-based and queue-based communication. Meanwhile, we improve the software extensibility and reusability by adopting multiple design patterns.The performance of runtime environment is one of the most important factors that constrain the transaction throughput in distributed transaction processing monitors. This thesis also discusses the optimization of runtime environment. OnceTX runtime environment uses thread pool technique to implement the scheduling and execution of services. Aiming at the distributed transaction application environment, this thesis introduces Leader/Followers pattern and put it into use to implement an effective thread pool, which is able to dispatch service request events efficiently, and is especially suitable for event-driven services. The experiment data show the performance improvement of service schedule and execution in runtime environment.