Study On Key Technologies Of Monitoring System For Supercomputer

As one of the most important achievements in the development of modern science and technology,the supercomputer plays more and more important role in national defense,science and technology,education,health care and other fields.From the early CRAY-1 vector machine to the currently fastest "Tianhe-2" supercomputer,we can draw the developing trend of supercomputer system.With the performance of supercomputer continues to improve,the system size is getting more and more large,the system architecture is also more and more complex,this put forwards an austere challenge for the system's monitoring and management.Based on the actual demand for domestic supercomputer system,the key technologies of the monitoring system are studied deeply in this paper.The main work and contributions list as following:1? Several typical supercomputers and its monitoring and management system are studied.Based on the in-depth study on the deficiencies among existing monitoring and management system,combined with the characteristics of small amount of date transfer,high real-time in monitoring and management system,a scalable hierarchical control architecture based on monitoring cluster is proposed in this paper.A parallel monitoring and testing protocol i-JTAG based on local area network & boundary scan,and a remote control protocol i-I2 C based on I2 C are then given.2? In view of the existing monitoring and management system with separated platform management and the rack management,and lower resource utilization rate,an integrated system management controller(SMC)combing the platform management and the rack management is designed.The logic design and engineering design of the monitoring management board are completed;the logic design of control module based on CPLD is completed.3? The scalable hierarchical control architecture based on monitoring cluster proposed in this paper has been applied in the actual supercomputer system,it can effectively monitor and manage the supercomputer system and thus to guarantee the system with high maintainability and safety,and also is the reassure of supercomputer system RAS technology.