Development of Big Earth Data Cloud Service Platform is underway to provide integrated service capabilities to provide unified computing and storage by breakthroughs in developing new methods of distributed computing resources, unified scheduling and analysis in order to achieve big data-driven scientific discovery and decision support.
The Cloud Service Platform has a hybrid architecture, integrates four subsystems, include HPC, big data cloud, data storage, and high-speed internal network, support customized data processing and release environment. The HPC subsystem is a 1.0 Petaflop (PF) compute cluster, interconnected with Mellanox HDR InfiniBand in a hybrid fat-tree topology. The big data cloud subsystem has more than10,000 cores, and can house up to 10,000virtual hosts.The Platform also features 35PB of performance storage (100GB/s aggregate), with spine-leaf network architecture. And through the unified service portal, the Cloud Service Platform provides scientific researchers with the services of scientific data exchange, big data processing and analysis, high-performance computing simulation, scientific research results release and display, important data storage and backup, and user-defined data processing platform rapid construction. The platform is opened from the data layer, which integrates the functions of HPC and big data cloud, avoids the trouble of multiple copies of the same batch of data, and greatly improves the efficiency of scientific research.
The Cloud Service Platform provides high-performance computing, scientific research data publishing and sharing, environment customization, online data analysis and mining and other services for scientific researchers through the cloud service portal. Especially, Earth Big Data mining analysis system EarthDataMiner, grid data engine DataBox and MPP database engine that supports GIS expansion and other self-developed software, etc.
Figure 1 Cloud Service Platform
