Research And Practice Of Large Scaled Distributed System's Service Objects Deployment And Runtime Monitoring

With the rapid development of network technology, the software platform is becoming from single computer to net, accordingly the computer system is also becoming from centralized system to distributed system. At the same time, because the distributed computing technology has been applied deeply and widely in many key domains, the distributed system has some new characters. For example, the system's scale has become larger and larger, the service objects have become more and more, and the relationship of objects becoming more and more complex. The new characters of distributed system bring on the system's management new problems, such as configure, deployment, runtime monitor, system maintain, and so on. The monitor and management on application service have become very important in large-scaled distributed system. So it is a research work of general and practical significance to solve the problem of management for large-scaled distributed system effectively, especially to satisfy the urgent demands of the service objects deployment consistency, and satisfy the demands of the service objects'state monitor immediateness.In the Background of the large-scaled distributed system's application management, aiming to the middleware application service object in distributed system, this article devises the ODMS (Service Object Deployment and Runtime Monitor System). The primary contributions of this article include:1. We design the architecture of ODMS, and divide it into five modules, such as Configure and Management Resources Module, Deployment Service Objects Module, Runtime Monitor Module, Object's Ref Management Module and Dynamic Maintenance Module. We place emphasis on the architecture and the characters of Deployment Service Objects Module and Runtime Monitor Module. Then we design the principle of the consistency about service object deployment in distributed system, and the mechanism of agent-based runtime monitor.2. We research on the technologies of the service objects consistent deployment in distributed system. We introduce some basic descriptors about the deployment consistency, and then put forward two constraints: the consistency between object and server node, the service object's version consistency between different server nodes. After this, we discuss the technology of deploying plan consistency maintenance when server's configure change or object updates. At last, we design the police of object transfer.3. We research on the deployment technology of service objects with dependence. We put forward the judgment algorithm of the deploy plan's dependence, and design the procedure of service objects having dependent relationship. 4. We research on the key technologies about the system service objects runtime monitor, and we place emphasis on the technology about object state management. We also analyze the safety on object state changing in real.5. We implement the ODMS based on the massive transaction processing middleware platform StarTPMonitor. The ODMS has been greatly used in massive date management platment.