Back to Table of contents

Primeur weekly 2013-12-09

The Cloud

EDF chooses Bull and Osiatis to deploy and run the world's biggest private Cloud-based infrastructure with Microsoft technology ...

Datapipe meets enterprise demand for HPC Cloud; guaranteed performance storage fueled by SolidFire ...

Researchers at Massachusetts Institute of Technology, the University of Texas at San Antonio and Notre Dame select OpenStack-based Hybrid Cloud to build federated Cloud service for research and education ...

IBM expands industry-specific Cloud Consulting Services ...

Helping clients securely and reliably migrate Cloud data ...

IBM, Dexia and several major financial institutions across Europe agree to a 1.3 billion US dollar sourcing deal for IBM to build and manage their IT ...

Türk Telekom Group signs agreement with HP to put Cloud services at customers' fingertips ...

EuroFlash

RenewIT: EC funded initiative to increase use of renewable energy in data centres ...

The HPC Advisory Council and the International Supercomputing Conference announce university teams for HPCAC-ISC 2014 Student Cluster Challenge ...

New supercomputer for the German Meteorological Service ...

Supporting European Early Career Researchers & Scientists working with Data participation to the 3rd RDA Plenary Meeting ...

PRACE Summer of HPC awards best students ...

PRACE Scientific and Industrial Conference 2014 HPC for Innovation: when Science meets Industry ...

Data centres can be cooled down in environmentally friendly, energy efficient ways ...

PRACE and XSEDE Call for Requests for Support ...

Bull helps organisations manage their data centres in the face of new energy issues ...

IBM and business partner Bechtle providing 6,100 System x and Flex Systems servers to European Union organisations ...

USFlash

Cycle Computing strengthens management team to meet growth demands ...

Supermicro 1U 4x GPU SuperServer in Green500 #1 ranked TSUBAME-KFC supercomputer from Tokyo Institute of Technology GSIC ...

NCKU develops world's first switchless cluster supercomputer ...

ViaWest supports Intel's TOP500 supercomputer project ...

Brocade 100 GbE core routers provide high-performance network to advance global research ...

Oracle releases Oracle Linux 6.5 ...

NCKU develops world's first switchless cluster supercomputer


26 Nov 2013 Tainan - The National Cheng Kung University (NCKU) Supercomputing Research Center (RSC) has built the world's first ever switchless cluster computing. Known also as the "CK-Star", this computing connects eight computers without switch control, thus breaking Intel's performance record, the university revealed recently.

Also, the single supercomputer, jointly built by NCKU RSC and Gigabyte, has also been recorded to reach 3.7 teraflops (3.7 trillion floating-point operations) in terms of computer performance.

Known as the "GS-R22PHL", this single supercomputer is currently the highest performing single supercomputer to date, thus setting a world record in the field of supercomputers and a new milestone in the history of computing.

Together with the computer systems provided by Acer, the "CK-Star" is jointly built by NCKU RSC Director Dr. Chi-Chuan Hwang and Mainland China's National Supercomputing Center in Jinan (NSCCJN) "Taishan Scholar" Dr. Yuefan Deng, who is also a Distinguished Professor at the State University of New York (SUNY).

On the other hand, the GS-R22PHL single supercomputer is a joint collaboration between NCKU and Gigabyte.

Both record-breaking research were presented at Yue-sheng Lecture Hall of NCKU Department of Engineering Science on November 18 and attended by NCKU Vice President Chih-Chin Ho, NCKU Computer and Network Center Director Jung-Hsien Chiang, and many other researchers and scholars.

NCKU RSC Director Dr. Chi-Chuan Hwang added that CK-Star is the most unique switchless cluster supercomputing as compared with traditional cluster supercomputing which requires switch control that controls interaction between nodes.

He explained that the weakness of the traditional one is that switches will become performance bottlenecks when the number of nodes is large.

Besides, switches also consume a large amount of power, estimated to be 50% of the total power consumption. Hence, these problems can be overcome with the introduction of the CK-Star which allows the unrestricted expansion of supercomputers.

CK-Star also integrates the Intel Xeon Phi coprocessor and has produced a record-breaking achievement in computer performance at 80.2% for four nodes and 77.3% for eight nodes, beating Intel's previous world record of 79.6% for four nodes and 76.1% for eight nodes.

"CK-Star has an innovative structure and a much enhanced performance. It has overcome the limitations of switches and solve the problem of high power dissipation through switches", Dr. Yuefan Deng explained.

Dr. Deng added that the network topology of CK-Star can also be adjusted based on application requirements to achieve the desired performance, thus indicating that the high-performance, flexible framework has opened a new route which shows high research value and commercial potential to high-performance computing systems in the future.

With the completion of Taiwan's first mainframe computer certification centre by NCKU RSC, the centre presented the outcome of its first research, which is the development and certification of the single supercomputer with Gigabyte.

Relying on the company's experience in product development and manufacturing techniques of Gigabyte, the GS-R22PHL single supercomputer with eight-slot universal graphics has been successfully developed.

Through the verification process carried out by the certification centre, the GS-R22PHL single supercomputer, added with the Intel Xeon Phi coprocessor, achieved 83.3% computer performance, beating Intel's previous world record of 79.8%. The computer performance of GS-R22PHL also maintains at around 78.2% when two coprocessors were used.

"Before the invention of Gigabyte's GS-R22PHL, ten or more servers were required to achieve similar computer performance", Dr. Hwang said while noting that this invention has written a new chapter in the history of computers.

He also stressed that the development of this single supercomputer with Gigabyte will strategically focus on software information and high-density computer systems and target the high-performance server market due to the high market potential and value of such computers in the commercial field.

Dr. Hwang also revealed that based on the foundations laid by these two major breakthrough research achievements, NCKU will focus on software integration and effective use of resources together with Acer whereas the NCKU-Gigabyte collaboration will emphasize more on supercomputer hardware set-up, design and development.

Meanwhile, NCKU RSC has also signed a memorandum of understanding (MoU) with K.E. & Kingstone Co. Ltd. to conduct "Supercomputer Workshops" which will include a series of supercomputer-related training courses and will be lectured by professors from NCKU RSC at K.E. & Kingstone's education and training centre.
Source: National Cheng Kung University - NCKU

Back to Table of contents

Primeur weekly 2013-12-09

The Cloud

EDF chooses Bull and Osiatis to deploy and run the world's biggest private Cloud-based infrastructure with Microsoft technology ...

Datapipe meets enterprise demand for HPC Cloud; guaranteed performance storage fueled by SolidFire ...

Researchers at Massachusetts Institute of Technology, the University of Texas at San Antonio and Notre Dame select OpenStack-based Hybrid Cloud to build federated Cloud service for research and education ...

IBM expands industry-specific Cloud Consulting Services ...

Helping clients securely and reliably migrate Cloud data ...

IBM, Dexia and several major financial institutions across Europe agree to a 1.3 billion US dollar sourcing deal for IBM to build and manage their IT ...

Türk Telekom Group signs agreement with HP to put Cloud services at customers' fingertips ...

EuroFlash

RenewIT: EC funded initiative to increase use of renewable energy in data centres ...

The HPC Advisory Council and the International Supercomputing Conference announce university teams for HPCAC-ISC 2014 Student Cluster Challenge ...

New supercomputer for the German Meteorological Service ...

Supporting European Early Career Researchers & Scientists working with Data participation to the 3rd RDA Plenary Meeting ...

PRACE Summer of HPC awards best students ...

PRACE Scientific and Industrial Conference 2014 HPC for Innovation: when Science meets Industry ...

Data centres can be cooled down in environmentally friendly, energy efficient ways ...

PRACE and XSEDE Call for Requests for Support ...

Bull helps organisations manage their data centres in the face of new energy issues ...

IBM and business partner Bechtle providing 6,100 System x and Flex Systems servers to European Union organisations ...

USFlash

Cycle Computing strengthens management team to meet growth demands ...

Supermicro 1U 4x GPU SuperServer in Green500 #1 ranked TSUBAME-KFC supercomputer from Tokyo Institute of Technology GSIC ...

NCKU develops world's first switchless cluster supercomputer ...

ViaWest supports Intel's TOP500 supercomputer project ...

Brocade 100 GbE core routers provide high-performance network to advance global research ...

Oracle releases Oracle Linux 6.5 ...