Cluster of "AL-100"

NPO Saturn is the leading engine-building corporation specializing in development and production of turbine jets. "AL-100" is not the first experience of Saturn in the field of high-performance computing. Implementation of the supercomputer brought into operation in 2005 allowed design divisions of the enterprise to move to the new level of analytical design based on use of numerical methods at all development stages of turbine jets.

The modern market of turbine jets demands from the developer of significant reduction of terms and costs during creation of a new product. For accomplishment of this task it is necessary to carry out the enormous amount of engineering calculations for the directions of gas dynamics, durability, acoustics. And in general it is difficult to revaluate modeling of a digital experiment on the supercomputer as it significantly reduces the number of expensive natural tests of products. Today with implementation of cluster technologies of NGO moved to qualitatively new level of accomplishment of design works, and with input of this supercomputer expanded the opportunities. As a result the development cycle of the gas turbine unit managed to be reduced three times, to four-five years against twelve-fifteen in 70 — 80kh years, and quantity of prototypes — five times, from fifty to ten. Besides, use of high-performance computing allows to cut significantly fuel consumption and noisiness of the developed engine that is also the important parameter influencing products konkurentospobnost.

However by the end of 2006 when branches in Moscow and Perm were connected to the supercomputer, loading of computing powers of a cluster came very close to 100%. Besides the volume of calculations increased by 160 times. And in the middle 2007go the decision on creation of the new supercomputer which had to bypass on performance already being available more than by 20 times was made. The design service became the customer of such project again. And the contractor, as well as for the first time, selected the CROC company which proposed the solution based on technologies of IBM.

Technical details

General architecture of a computing cluster of "AL-100"

The created solution represents a high-performance cluster based on the solution IBM System Cluster 1350 with a peak capacity of 14.3 TFLOPS. The HS21 servers edges based on quad-core Intel Xeon processors integrated by high-speed network DDR Infiniband (see the drawing) became a basis for a computing field. The rated field of a cluster includes 1344 main cores and 1344 Gbytes of RAM. Software of IBM Cluster Systems Management is responsible for management of a cluster. For management of computing tasks the IBM Tivoli Workload Scheduler LoadLeveler system is a part of cluster software.

The operability of the computer system is provided with the engineering infrastructure including no break power supply systems, precision conditioning and automatic gas fire fighting. The basis for creation of infrastructure selected the complete solution APC InfraStruXure. The solution includes the system of power supply with a power of 250 KW, air cooling with hot and cold corridors and also water cooling in conditioners. As a result it was succeeded to achieve more high density of placement of the equipment, it is essential to reduce costs for the electric power ("AL-100" consumes only 72 kW) and to reduce noise level. All complex is integrated by the systems of the centralized monitoring and automation.

For launch this supercomputer was the most highly productive in the industry of Russia and the CIS, takes the fifth place in the list of the Top 50 supercomputers of the CIS and entered ten the most powerful in industrial enterprises of Europe. According to forecasts of capacities of the supercomputer should be enough for two-three years. And 50 — 60% of its resources are supposed to be given for development of the engine for the fighter of fifth generation.

Costs for creation of the AL-100 supercomputer made about 100 million rubles. 30 million left them on development by own efforts of software for engineering calculations (solving of tasks of gas dynamics and durability, optimization of construction of the engine), and 70 million — directly on the equipment and the system software, reorganization of premises, etc. At the same time according to managers of Saturn on engineering infrastructure about 40% of all project expenses left.

Completely it is supposed to pay back these costs quickly enough. The first supercomputer of NGO paid for itself already later half a year. In the last three years in Saturn passed to virtual design. The SaM146 engines and AL-55I which are nowadays passing flight tests were quite so developed.

The systems of such level in world practice have own names. In NPO Saturn competition among workers on the best name of the new supercomputer was announced. From several hundred offers the winner recognizes the name "AL-100" — in honor of century since the birth of the founder of the company, the outstanding scientist and the designer Arkhip Mikhaylovich Lyulki to whose engines the IS SCARLET brand was traditionally appropriated.

General architecture of a computing cluster of "AL-100"

Links

• Supercomputer