Next to HPC development, T-Platforms has also a software and services department with a strong vision of HPC solutions towards customers with regard to software and services.

Andrey Slepuhin told the audience that the latest project finished in 2008 was the installation of a supercomputer at Moscow State University. This machine has 625 blades and includes 1250 quad-core Intel Xeon E5472 processors, and achieves a 60 Tflops peak performance.

The T-Blade system for Moscow State University is the most dense blade solution and the first blade to support Intel 5400 chipset guaranteeing up to 30% better real application performance.

The standard PCI-Express expansion slot is compatible with any standard interconnect and performs up to 435 Wt per blade maximum.

In 2009, the company is developing the second generation of T-Blade with support for Intel Nehalem and AMD Istanbul CPUs improved management. This second generation is being designed from scratch with a very big density including 64 Intel Xeon Nehalem series X5500 in 7U. It provides 18 Tflops per standard 19" 42U rack.

It has an integrated MicroSD slot for boot flash support. Future development will include AMD socket G34 and GPGPU blades. HPC-targeted networks will have global barriers and global interrupts. It is a scalable system with ConnectX QDR Infiniband chip and has up to 3GB DDR3-1333 SDRAM per core.

The system provides an integrated management node including compute nodes remote control, diskless compute nodes boot, chassis fans and PSU monitoring and control and integrated switches control. As for the integrated communications, there are 2 x 36 ports QDR IB switches and a Gigabit Ethernet switch.

Another type of systems which was presented by Andrey Slepuhin are the PowerXCell 8i based products, developed in close partnership with IBM and to IBM's order. The 1U server has 2x8 core PowerXCell 8i processors, up to 32 GB RAM, standard PCI-Express extension slots and a 410 GFlops single processor.

The Linux HPC OS distribution has a performance optimization for kernel and system libraries; a new memory management for transparent usage of huge pages; and a transparent support of OpenMPI.

Support of T-Platforms T-Blade II features global barriers, global interrupts and global clock extender with optimized DHCP and TFTP server and support up to 25.000 nodes.

The integrated cluster monitoring and management solution has no single point of failure; comprises a modular architecture which can be easily configured without performance penalty and includes up to 25.000 nodes that can be monitored from a single server, and up to 300 metrics from 1 node with 1 sample.

The AutoParallelizer is a tool that looks for parallelizable parts of source code of applications written in a sequential manner and language. The AutoParallelizer uses libraries that support OpenMP standard and are currently integrated with GCC framework. It has the benefit to work with every language, as explained by Andrey Slepuhin.