Support Enterprise Applications Distributed File Cluster System

To deal the mass enterprise unstructured data and the limitation of database, the requirement to a universal file service platform that has well performance and well reliability becomes more and more urgent. This dissertation introduces the research and implementation of a distributed file service group system. After having described the way of software architecture designing and analyzed several typical architectures, we decide to use distributed system as our architecture's main technology. To support many application software architectures and the natural organization of enterprise, we design our system's architecture as a distributed system named federated architecture. This architecture has the ability of well performance, mass data dealing, well reliability, good expansibility, good integration and high security. The system should take some way to heighten system performance. We use part-level cache to heighten one-server. It uses full cache to deal small files and uses part cache to deal big files. So it heightens response time and performance in one server. The system uses automation file's location optimization and load balancing to heighten multi-server. It can heighten the whole system performance by the collaboration of all servers. Besides the system performance, the system reliability is the main factor we should consider. But the reliability is opposite with performance. Heightening reliability will influence performance. So we balance the system performance and reliability, we furthest heighten the system reliability under the permission of performance. We introduce some ways in implementation: 2 data replicas twin-circle storage, consistency guarantee, long-distance backup etc. At last, we introduce three systems based on the distributed file service group system platform: product data management system, knowledge management system and workflow management system.