Preliminary Research On Distributed Collaborative High-performance Computing Framework For Spatial Information

The space weather in the Sun to Earth space directly affects the national defense, aerospace, economic development and human life. Therefore the monitoring, researching and forecasting of the space weather are of great significance. Nowadays how to process and analysis the Sun to Earth spatial information obtained by monitoring, and finally achieves accurate space weather forecasting, is one of the hot spots of space science application research. Adopting the scientific computing Common Component Architecture introduced by America recently, service oriented pattern, high-performance computing, and distributed computing theory and technology, this paper brings forward a Distributed Collaborative High-performance Computing Framework for Spatial Information, DCHF-SI. It supports component encapsulation for the space physic model, construction and management of simulation application, interoperability of models, collaborative computing, distributed fault-tolerant, computing steering and interactive visualization. It is capable of implements multi-physics loose coupling simulation on internet based on peer component and service oriented pattern, distributed collaborative high-performance computing of spatial information, and ultimately achieves space weather forecasting. The main research results of this paper are classified into four categories:(1) DCHF-SI architecture is proposed. The layered architecture model DCHF-SI-LM based on the CCA architecture, expended CCA specification, component encapsulation pattern, scientific component unified description/ discovery/ integration protocol SC-UDDI, and the DCHF-SI running mechanism are proposed.(2) Theory of implementation the basic framework for coupling interaction is proposed and verified. Based on CCA specification, the implementation strategy of its three core parts are designed including a framework driver, a builder and a coupling interaction mechanism. A parallel coupled interaction model PCIM is refined. Based on PCIM, a parallel coupling interaction mechanism PRMI + + is presented which is the keystone study content, and better results are obtained.(3) The core framework service components are researched deeply. The checkpointing and rollback recovery strategy in distributed parallel environment named DPCRR and a distributed fault-tolerant service component DSFT which is based on DPCRR are proposed; Based on CCA specification, a remote parallel visualization component PRView which provides real-time interactive visualization services is designed; Based on CCA specification, a distributed computing steering component DCSteering which provides computing steering services is designed. This research lays a theoretheal foundation of the further study.(4) A cross-platform graphical user interface DHCF-SI-GUI is designed and built. It encapsulates the complexity of the underlying framework, and could enable users to assemble, manage and steer application only just through mouse clicking, dragging and dropping. Based on DHCF-SI-GUI, a real-time online three-dimensional visualization of Coronal Mass Ejection is designed and implemented.