case study
Construcor: transforming scientific research with Machine Learning, AI and Virtuozzo cloud
Constructor is a global science, education, and technology organization that aims to provide the critical knowledge, solutions, and services to help solve the most pressing global issues. Constructor offers a Platform and Infrastructure for research with consulting and Machine Intelligence-based applications for academia, sports, and business.
The company has strong expertise in MI and data science in various application areas, including scientific research, business analytics, professional sports, driverless mobility, robotics, and others.
The Constructor platform simplifies and accelerates the research lifecycle through AI-based computational modeling.
It solves a problem that most research teams face today, as they need increasing compute power to model and analyze large datasets. For example, that compute is usually deployed and configured on laboratory servers or inside a university data center.
As a result, scientists must also become developers, infrastructure specialists, and DevOps practitioners to get their work done.
“What we provide is an advanced research collaboration platform, with machine intelligence, tightly integrated with high-performance virtualized infrastructure,” says Andrey Omelyanchuk, VP of Engineering at Constructor.
“That way, research teams can focus on research, not IT. They can get the tools, and IT resources they need with a few clicks.”
The team is already bringing various internal Constructor projects onto the platform. Ultimately, Constructor’ goal is to offer its research platform as a service to commercial and academic customers worldwide.
Flexible CPU & GPU virtualization
The IT challenge particularly tricky for researchers is the wide range of infrastructure types they need, not just for different projects but also for different stages of each project. It is all the more accurate as science teams look to use more efficient Machine Learning and AI-based computational modeling for their projects.
One example is a materials project involving Professor Sir Konstantin Novoselov, an advisor to Constructor team. Professor Novoselov is a renowned materials scientist and was a joint winner of the 2010 Nobel Prize in Physics for co-discovering graphene. The new project involves computational modeling of new materials and creating machine learning models that can predict new materials’ interesting chemical and physical properties without requiring large amounts of high-performance computing.
The process does require a lot of compute, as the initial modeling of the material requires difficult DFT (Discrete Fourier Transform) algorithms. Once some interesting material properties have been calculated, the next stage is to train machine learning models to predict these properties without needing much compute resources – typically using neural networks running on GPUs.
Konstantin Novoselov
Nobel prize winner, professor of physics at the National University of Singapore
The Virtuozzo cloud platform
The Constructor platform runs on Virtuozzo Hybrid Cloud, the public cloud powered by Virtuozzo Hybrid Infrastructure, and on Virtuozzo Hybrid Infrastructure deployed in Constructor’ own data center. Virtuozzo provides high-performance Kubernetes clusters and S3 storage, and GPU-powered virtual machines for the Constructor research platform.
“Virtuozzo helps enormously by making it easy to switch between CPU and GPU virtualization and by providing this flexible mix of infrastructure in an easy-to-manage platform,” Andrey Omelyanchuk says.
The Constructor application runs on a mixture of traditional virtual machines and scalable Kubernetes containers. Constructor has integrated Virtuozzo within its HPC cluster to simplify the end-to-end process, with its platform providing a simple and intuitive front end.
A hybrid research cloud
Since day one, a hybrid of remote and on-premises deployment has been part of Constructor plan. While the public cloud model provides the easiest way to get started, it does not suit every use case – especially when researchers must comply with national or industry regulations about data protection and data must remain within an organization or a particular region.
As a result, Constructor also provides an on-premises solution, allowing its research platform runs on top of a private Virtuozzo cluster deployed in a customer’s lab or data center. Virtuozzo’s OpenStack API makes it easy for research teams to combine on-premises and cloud in hybrid configurations.
“Virtuozzo fully supports the OpenStack API, which means we can offer the flexibility for customers to combine on-premises and remote resources,” Andrey says.
“Thanks to the Virtuozzo platform, we can have those discussions with new customers and offer tight integration between the two. Virtuozzo clusters are also very stable for Kubernetes and virtual machines. Virtuozzo’s Kubernetes scalability and mesh configuration, with an easy-to-use management UI, is appreciated by our DevOps team.”
The Constructor platform is cloud-agnostic, although the team considers Virtuozzo to offer the best price/performance ratio for most use cases. It enables any scientist to create an account at constructor.tech, create their project, and start using the compute and GPU resources they need. However, if a team prefers to use one of the large public cloud providers or run Constructor on VMware or an in-house OpenStack cloud, they can.
Working with the Virtuozzo team
The Constructor team has worked closely with Virtuozzo during the development of its platform. According to Andrey, Virtuozzo has also been impressive from an engineering perspective, helping build more predictable installations, test infrastructure performance, and identify bottlenecks in the infrastructure topology or application code.
“There are many benefits to working with Virtuozzo. It’s a great relationship, and that’s important because we know we can discuss any roadmap changes to the Virtuozzo platform that will support the development of our platform,” he says.
“The Virtuozzo support team has been great, helping us optimize GPU virtualization inside our container-based environments, for example. The team is open to discussing and brainstorming solutions, which is crucial in our mission to build the most effective research platform.”
With Virtuozzo, Constructor is on track to make complex research projects much easier for scientists worldwide. Learn more at constructor.tech/products/research-platform.
Andrey Omelyanchuk
VP of Engineering at Constructor
Ready to see what the power of Virtuozzo can do for you?
Request your demo today.
