14 DAY TRIAL // IBM announced recently that the company has reached an agreement with the High Energy Accelerator Research Organization of Japan to develop for the group the "KEK Central Computer System," a server and storage solution that will be able to handle the petabytes of experiment data managed by the organization.
The system is expected to go live in February of 2012 and will be used by scientists for analyzing data generated from experiments in elementary particle and atomic nuclear research, as well as data from synchrotron radiation, neutron and muon research.
In order to handle these tasks the KEK Central Computer System will consist of more than 370 IBM iDataPlex systems that are based on Intel's Xeon 5600 series processors with up to six computing cores.
Each one of these systems can pack two such processors as well as up to 192GB of system memory, and are connected through each other using a high-speed Infiniband link.
Furthermore, IBM states its iDataPlex solutions were optimized in order to double the number of servers that can be run in a single rack, while also saving up to 40% more energy than other similar solutions.
For the Japan computer system, IBM will develop a special storage environment that consists of the IBM General Parallel File System (GPFS), which provides high-speed processing, and the High Performance Storage System (HPSS).
These were needed in order to handle the petabytes of experiment data gathered by the organizations, and the GPFS disk system will provide up to 7 petabytes of data capacity, while the HPSS system will provide up to 16 petabytes of effective data capacity.
Until now, data gathered through experiments at the organization's high intensity proton accelerator facility, "J-PARC," and also at its collider accelerator, "KEKB," were analyzed using separate systems installed at the respective facilities.
In addition to the scientific work, the new IBM system and storage solution also be used by domestic and international collaborative research institutions for sharing resources and data.