HPC HUB

HPC HUB is the Russian developer of the virtual supercomputer platform of the same name, on the basis of which HPC as a service/HPCaaS has been provided since 2016.

The founders of the company are graduates of MIPT of different years Wilhelm Bitner, who worked at Parallels and the Russian office of IBM, Evgeny Protasenko, specializing in investments, Andrei Nikolaev, who previously headed the parallel computing laboratory at MIPT, and Denis Lunev, lecturer at MIPT, holder of more than 15 patents in the field of virtualization and more than 15 years of experience at Parallels.

The idea of ​ ​ creating the project belongs to Wilhelm Bitner. At the company, he deals with operational issues. Evgeny Protasenko is responsible for issues related to strategy and investments. Colleagues on the HPC HUB were familiar long before the project, since the institute times, when they worked together and were friends, they told TAdviser.

HPC Hub is a resident of Skolkovo.

History of creation and development of the project

2017: Attracting investments and receiving a grant from Skolkovo

In the summer of 2017, HPC Hub was able to attract the next round of private investment from a group of investors, which was partially used to receive a grant Skolkovo"" (the grant was taken to finalize certain key components of the technology).

For HPC Hub customers, one base site is available at the facilities of Rosatom's subsidiary in Sarov with a capacity of over 35 Tflops. In the near future, another one will appear at the RKS facilities. At the same time, the company is simultaneously developing the direction of introducing private-cloud solutions and solutions for building hybrid clouds for large customers who own their own infrastructure.

2016: Market Entry

In December 2016, HPC HUB announced the release of the beta version of the on-demand supercomputer cloud service and the start of work with a wide pool of customers. The service provides access to a virtual supercomputer: a set of high-performance servers combined into a single computing system by the Infiniband low latency network and managed by cluster software. This makes HPC HUB different from cloud services like Amazon Web Services, Microsoft Azure and Google Cloud, which provide mainly access to isolated virtual machines, the company says. The service also includes pre-installed accounting software, including commercial software.

The service is based on the virtualization platform of its own development HPC HUB. The core of the service is deployed in the city of Sarov on the basis of computing resources rented from one of Rosatom's subsidiaries, the company told TAdviser. At the initial stage, HPC HUB leased about 1600 cores of the computing cluster.

Immigrants from MIPT and Parallels launched a cloud supercomputer as a service

The cloud infrastructure of the HPC HUB can be used to solve problems such as field modeling and seismic data processing, natural process modeling, data processing and modeling in bioinformatics, development of innovative drugs in pharmaceuticals, industrial modeling tasks, as well as for network rendering and video data processing.

By the time HPC HUB entered the market for several years in Russia, there are organizations providing high-performance computing resources for rent. Such services are provided, for example, by T-Services, which is part of the T-Platform s. In addition to computing power, the company also leases application software for calculations, carries out resource-intensive calculations at its own capacities, etc. Rosatom structures also provide remote access to their own computing clusters. In the field of science and education, some universities provide remote access to their supercomputers, for example, Moscow State University, SUSU, and the Samara University Supercomputer Center.

HPC HUB notes that existing providers, unlike their company, provide access to hardware without virtualization, which means "cutting" allocated resources manually, without automation. This means large overheads, and, as a result, limiting the use of the service for small organizations, says Yevgeny Protasenko. Among the differences between its service and similar ones without virtualization, HPC HUB calls the ability to freely install software on a virtual supercomputer (root rights are granted to the client) and the lack of the need to coordinate changes with the admin, the ability to save a configured environment, create recovery points and transfer the configured environment to another cluster if necessary.

Another difference, they say in the company, is the ability to rent a minimum number of nodes, if desired - for a very short period and quickly gain access to them. This opens access to the service not only for large, but also for small organizations.

HPC HUB sees companies that already provide such services, but without virtualization, as potential partners or customers rather than competitors, Protasenko told TAdviser. They can install the HPC HUB virtual platform on their supercomputers and, based on it, provide access to their computing resources to their customers to simplify resource utilization and management, he explains.

Sales of not the end service, but their virtualization platform in HPC HUB are considered as one of the possible business models. In it, customers can, for example, be owners of supercomputer capacities, whose resources are idle. By analogy with Rosatom, they can give part of their resources for the provision of HPCaaS services. Also, owners of supercomputers can buy a platform to deploy a private cloud on it.

Evgeny Protasenko noted in a conversation with TAdviser that the company is just entering the market and has not yet decided on the business models in which it will continue to work. Different options are being considered.

It is quite possible that the most profitable path for us is to be a technological vendor who sells his technology to the sites and allows them to utilize their own capacities more efficiently, he argues.

The first customers of HPC HUB were a number of organizations using the service to solve problems in the field of bioinformatics, oilfield service, rendering, the company told TAdviser. Oilfield services companies are the main focus for HPC HUB. One of the existing clients in this industry is NPO Soyuzneftegazservice.

According to Yevgeny Protasenko, at the time of entering the market, the company lives on the proceeds that it already receives. The sales that the company expects to make in the near future should cover all previously invested funds in the project.

According to HPC HUB, the volume of the Russian market for access to high-performance resources on demand, taking into account various factors, may vary from 1 to 5 billion rubles.

2015: Investments from Dmitry Mikhailov and IIDF

In 2015, the first investor Dmitry Mikhailov, CEO of EG Capital Partners, joined the project. According to TAdviser, the company received about 12 million rubles from the investor.

Subsequently, HPC HUB received a little more than 2 million rubles from the IIDF.

2011-2012: Development begins

The idea to create a virtualization platform in the field of HPC and transfer all the advantages of clouds to this area arose at the technical backbone of the team back in 2011, Evgeny Protasenko told TAdviser. However, then it was not clear how technically possible, because technologically the task is different from creating a platform for server virtualization.

When the team experimentally found out that the problem was solvable and created a prototype in 2012, the question arose of what would happen to the HPC on-demand market, since it was new for Russia. The market is very narrow, inconspicuous and the project did not arouse interest among investors at that time, says Protasenko. Thus, the development remained for several more years for the team something of a hobby.

A couple of years later, the niche of this market was outlined, it became clear that the demand for high-performance computing services exists not only in the segment of large organizations, but also in small companies.

The development of the service was carried out on the basis of the supercomputer of the Kurchatov Institute.