Cooperation between CAS and Local Authorities
Dawning-6000 PetaFLOPS System Passed Acceptance Check
On Feb. 24th, "Dawning-6000 PetaFLOPS High-Productivity System", a project jointly undertaken by the Institute of Computing Technology, CAS, the Dawning Information Industry Co. Ltd. and several other research institutes passed the acceptance check. In November 2011, the Dawning-6000 system was deployed successfully at the National Supercomputing Center in Shenzhen (NSCS). It is comprised of 258 racks, 5600 Dawning CB60 blade servers (each equipped with two Xeon X5650 microprocessors), 2560 Nvidia Fermi C2050 GPGPUs and 128 Dawning A840 servers. The total memory capacity of this system is 226 TeraBytes. It is interconnected through multiple networks, including an InfiniBand QDR 4X (40Gbps) high-speed network and two sets of Gigabit Ethernet. It symbolized breakthroughs on key technologies for PetaFLOPS computing based on CPU-GPU heterogeneous architecture and cloud computing management software for high performance computing environment. The back-up team of Dawning-6000 also conducted extensive research for the design and delivery of Super-Dragon system based on HPP architecture, such as HPP architecture, HPP system controller, node operating system, and collective communication speedup network. The heterogeneous architecture of X86 and Loongson CPUs were explored. Progress has been achieved on the interoperability of Loongson and X86 software stacks, which paved the way for homemade CPUs to support mainstream commercial softwares. After being put into production usage in NSCS, Dawning-6000 has provided computing services to nearly 200 research institutes and users.