Real-time Multi-task Cluster Management System

With the great development of computer technologies, although the performance and reliability of single computer is improved quickly, but there are still many computing tasks can't be achieved by single computer. Cluster technology can connect many distributed physical devices through network and provides reasonably computing ability compared to super computer. On the premise of satisfying system real-time requirement, real-time multi-tasks cluster management system is able to manage and schedule software and hardware resources reasonably, and it can also improve system's utilization and throughput.This thesis discussed the basic concept of cluster, and analyzed several existing cluster system. Then we studied emphatically the architecture and some key technologies of real-time multi-task cluster management system. The main works of the thesis are as follows:1. The features and demand of real-time multi-task cluster management system are analyzed, and a three-level architecture based on C/S is established;2. The method of partition is proposed. Using partition, not only ensures that a job is executed in one special partition, avoiding the burdensome exchange information among the whole system, and also enhances the security effectively;3. A task scheduling strategy is implemented. Each task is scheduled into a node which has the least task executing time and the load is correspondingly lower;4. The system adopts the active, local, and dynamic load balancing strategy. It avoids to broadcasting load in the whole system and can reducing the communication of the network;5. Based on a business cluster system Symphony, a real-time multi-task cluster management system is also constructed, which is used in ARINC429 avionics bus data acquire system. And the performance of whole system is analyzed in detail.