SberCloud increases cloud platform opportunities for work with AI using oneAPI

May 20, 2021

At SmartDev on May 20, 2021, Sber announced that it would be enhancing the capacity of SberCloud ML Space, a cloud platform for full cycle development and implementation of AI services, offering tools and resources for the creation, training, and deployment of machine learning models, from quick connection to data sources to automatic deployment of models in SberCloud.

ML Space is a cloud service for organizing distributed learning using Intel® Xeon® Scalable Processors with built-in AI acceleration. Its architecture is based on SberCloud’s supercomputer, Christofari, which has a benchmark power of 6.7 petaFLOPS and is ranked 40th among the top 500 supercomputers in the world.

The strengthened capacity was achieved by adopting the open oneAPI cross-architecture programming model, allowing the developers to effectively exploit the performance and unique capabilities of various architectures, without having to adjust the code for each hardware platform. This gave them the freedom to choose the best equipment for each concrete task. oneAPI also supports known programming language (e.g., C, C++, Fortran, and Python) and general standards (e.g., MPI and OpenMP), ensuring interoperability and close alignment with existing code.

The SberCloud ML Space platform was created, on the one hand, to provide data scientists with the best tools for solving problems in machine learning and, on the other hand, to streamline and democratize the development and use of AI-based products as much as possible. The new Intel oneAPI Toolkits fit in with the ML Space ideology perfectly. Data scientists and ML Developers working on efficient, flexible, and economical processor architecture will be able to accelerate development and implementation of their AI products and improve their characteristics. I am certain that specialists will appreciate the possibilities our cloud platform presents.

David Rafalovsky

CTO, Sberbank Group; Executive Vice President, Head of Technology, Sber

ML Space combines popular big data tools – Jupyter Notebook and Jupyter Lab – with productivity tools – Intel oneAPI Toolkits. It is based on a modular architecture, allowing users to add new features. For one year following this announcement, anyone will be able to register and receive test access to the SberCloud ML Space platform, Intel® oneAPI Toolkits, and Intel® processor-based servers.

As the world moves toward the XPU era based on a combination of architectures, developers will require new levels of abstraction to ensure the efficiency of the underlying hardware. We are delighted that Intel® oneAPI Toolkits will help SberCloud AI cloud platform users effectively create high-performance software for various architectures using a single codebase.

Raja Koduri

Senior Vice President, Chief Architect, General Manager for Architecture, Graphics, and Software, Intel

Intel oneAPI Toolkits help developers effectively create, analyze, and optimize high-performance cross-architecture applications for various XPU: Intel processors, graphic processors, and FPGA.

These toolkits provide support for the oneAPI cross-architecture programming language Data Parallel C++ (DPC++) and over 40 software products: compilers, libraries, and migration, analysis, and debugging tools, which simplify the work of data processing applications.

The Environments module, one of the ML Space platform’s key elements, will receive the following Intel oneAPI toolkits:

  • Intel oneAPI Base Toolkit – a core set of tools and libraries for developing high-performance cross-architecture applications.
  • Intel oneAPI AI Analytics Toolkit gives data scientists, AI developers, and researchers familiar tools and frameworks to accelerate data science and analytics on Intel CPU and GPU architectures.
  • Intel oneAPI HPC Toolkit delivers what developers need to build and analyze high performance applications based on Fortran, OpenMP, and MPI, scaling them across new Intel processor-based systems and clusters. Together with the Base Toolkit, it provides everything that is required to develop high performance applications to solve science or engineering problems on shared or distributed memory systems.
  • Intel Distribution of OpenVINO Toolkit helps optimize, tune, and run comprehensive inference using a deep learning model optimizer and runtime and development tools.

About SberCloud

SberCloud (OOO Oblachnye Tekhnologii, or Cloud Technologies LLC) is the Sber ecosystem’s cloud platform. The company offers a wide range of infrastructure and platform cloud solutions, as well as AI tools based on the Christofari supercomputer, the highest performance computer cluster in Russia. SberCloud’s IT platforms and services are at the core of the Sber digital ecosystem and are also available to external clients such as companies and public sector organizations.