This presentation deals with the questions of HPC hybrid clusters construction on the
basis of HP typical solutions (S6500 chassis, servers of series HP SL390) and GPU
Nvidia Tesla. In the report there are listed examples of HPC solutions for customers in
the area of education, science and industry both for the elementary configuration, and
for a level configuration of PetaFlops performance are resulted. Peculiarities of
benchmarking and testing of similar decisions are discussed. The possibilities of using
Open Source software (Salome, OpenFOAM, Paraview) to solve the problems for
continuous mechanics at carrying out in cloud computing are discussed on the example
of a program ‘ University Cluster ’ (www.unicluster.ru) and operating web-laboratory
(www.unihub.ru).
