At the Krasnoyarsk state pedagogical university the supercomputer appeared

The company Open Technologies, the Russian system integrator, announced in December, 2012 project completion of creation of the hybrid supercomputer for the Krasnoyarsk State Pedagogical University (KSPU) of V.P. Astafyev.

The decision on purchase of the hybrid supercomputer was made by the management of the university within program implementation of strategic development. The high-performance complex was necessary for training of students in work with the CUDA platform and support of the solution of a number of scientific tasks. KGPU became the only educational institution in Krasnoyarsk which managed to win a grant on financing of this project.

Auction on creation of the computer system the company won Open Technologies. The system integrator could propose to higher education institution the solution optimized at cost allowing to solve the majority of educational and scientific problems of the university.

The task to create the computer system which, on the one hand, would meet current demands of the university and, with another was set for the contractor, during implementation did not require expensive upgrade of the providing subsystems of DPC. Also the possibility of further development of the supercomputer without change of the architecture put originally and ensuring protection of the enclosed investments was important.

As a basis for a complex the standard configuration of the hybrid ENTER 8/10 supercomputer (own development of the company Open Technologies) with an expanded part of monitoring and management was selected. Such approach allowed to minimize project cost and to win in the cost and terms of execution of works.

In work flow on the project by specialists of the company Open Technologies adaptation of the basic solution for an optimal solution of tasks of the customer was executed, work planning on a complex commissioning is made, installation and installation of hardware and program components is executed, integration of the supercomputer into engineering and information infrastructure of the university is performed.

"Implementing solution, offered by the company Open Technologies, allowed us to adjoin ranks of the universities using own supercomputers in the work. The offered architecture will give us the chance to apply supercomputer resources most effectively, and at emergence of requirement to growth - to perform it with the minimum costs and is absolutely transparent for users of a system", — Valery Beloshapkin, the project director of the program of strategic development of KGPU of V.P. Astafyev told.

Existence of own supercomputer will give to the university the chance to implement numerous advantages of hybrid computer systems. And considering that the number of the software packages focused on hybrid parallel systems as avalanche grows, the share of the algorithms effectively implementable on such systems increases, it is possible to speak, as the value of the computer system in life of KGPU of V.P. Astafyev, and together with it and a share of the found advantages, will only increase.

According to Anton Palmin, the regional representative of the company Open Technologies in Krasnoyarsk, "this project is bright demonstration of the fact that the supercomputer can be purchased at quite adequate price. We are always open for cooperation with the universities and intend to continue to develop the supercomputer industry in the SFD".

In more detail about technical aspect of the project:

• The KGPU supercomputer is constructed using the graphic processors of NVIDIA company allowing to increase significantly the capacity of the computing system at slight increase of cost and energy consumption in comparison with implementation only on central processors.
• Computing nodes of the supercomputer are executed in construct of high density with the built-in Infiniband FDR controllers allowing to build network of internodal interaction with the minimum delays and with capacity to 56 GB / c. A computing part based on graphic processors is implemented in the form of the stand-alone graphic modules NEXTIO vCORE Express 2075 allowing to place 4 GPU Tesla in 1U of mounting amount.
• Feature of the solution is that the used modular construct of hybrid computing nodes of a cluster allows to separate lifecycles of GPU and CPU of parts, having provided thereby their independent upgrade, and it is essential to simplify scheduled works. Such approach allows to approach also flexibly the solution of a problem of optimum allocation of sources of heat in a mounting cabinet, providing thereby its more effective branch. What will become especially relevant at increase in quantity of computing nodes of a cluster further.