NUDT and Canonical bring OpenStack to world’s fastest supercomputer
by Canonical on 14 May 2014
China’s National University of Defense Technology, NUDT, developers of the Tianhe-2 supercomputer, and Canonical, the organisation behind Ubuntu, today announce a collaboration to bring OpenStack to the world’s fastest supercomputer for high performance cloud environments.
The new collaboration with Canonical will enable Ubuntu Server, Ubuntu Openstack and Ubuntu’s orchestration tool, Juju, to run Tienhe2. Today, Ubuntu OpenStack is running on 256 high performance nodes and this will grow to over 6400 nodes in the coming months. The nodes will be available to Government departments in Guangdong province as well as other NUDT partners for analysis, census, and eGovernment applications.
Both OpenStack and Ubuntu’s orchestration tool, Juju, will run on Tianhe-2 to enable NUDT partners and affiliate to rapidly deploy and manage very high performance cloud environments. The Juju orchestration tool makes it easy to design, deploy, scale and manage cloud workloads in OpenStack environments. Workloads running on Tianhe-2 will enjoy higher inter-connect bandwidth and computing power for point heavy or memory intensive application.
Professor QingBo Wu at NUDT comments; “NUDT is a pioneer of technology, especially in the area of high performance. Tianhe-2, the world’s fastest supercomputer runs on Ubuntu Kylin and now with OpenStack and Ubuntu Juju, we are able to deliver high performance OpenStack.”
“To see the fastest supercomputer running OpenStack is already a beautiful thing,” said Mark Shuttleworth, founder of Ubuntu. “To see it running OpenStack with workloads orchestrated by Ubuntu Juju is incredibly powerful. We can’t wait to see it rolled out further.”
NUDT designed and built Tianhe-2, which runs on its own Kylin Cloud Linux operating system and has held the record for the world’s fastest supercomputer since 2013, having recorded results of Linpack Performance (Rmax)33,862.7 TFlop/s. The servers use Intel Xeon processors, Intel Xeon Phi co-processors and a 160Gb per second interconnect for super-fast data transfer between nodes.