Search
Close this search box.

Case Study

Örebro University

Making AI available and useful for all science

Share now

Proact Customer - Örebro University

Industry

Solution provided

To make AI available to scientists in every field, Örebro University added tools that make it easier to get started using its computational resource.

Initial situation

To meet the growing needs from AI workloads, the university procured a powerful, unified computing resource. The assignment to deliver this went to Proact, building on its solid experience from providing IT infrastructure and associated services to Örebro University. The university settled for a solution based on the ONTAP AI reference architecture, a computational server powered by thousands of GPU cores, combined with fast storage. It was possible to use and upgrade an existing NetApp storage solution previously delivered by Proact, resulting in a modern, while also cost-effective solution for the university’s needs.

Örebro University’s specialisation in AI was obviously a decisive factor behind the investment in the system. It immediately became useful for research areas which typically has a large appetite for computational power, such as imaging, robotics, and crypto analysis. But from the outset, the intent was also to make this resource
as widely used as possible, for areas and use cases that are new to AI and machine learning.

“We have requests from various departments and researchers in many areas who want to explore what AI and machine learning has to offer them. This interest spans from mathematics and physics to genetics and even criminology,” says Hans Jevrell, Enterprise Architect at Örebro University.

Collaboration to spread AI use

To help build skills and make AI more accessible, the university has also engaged in knowledge transfer, working with external parties such as the AI Impact Lab, a collaboration platform for the Örebro region. With its computational resource and research expertise, Örebro University also forms a node in the national AI  Sweden partnership.

AI research in Örebro previously had to rely on smaller, standalone computational servers, locally managed and operated by departments. The on-premise AI computing resource not only added loads of performance; because it is covered by data protection and access control managed by the university’s IT organisation, it also brought enhanced security. This is important for addressing the strict privacy and security requirements, as well as the ethical issues that arise with AI, not least when applied for medical and social sciences.

Proact’s mission was to support Örebro University in becoming an even sharper AI competence center. While the solution contains the latest in infrastructure for AI processing, it is cloud-enabled and gave the  opportunity to build on an existing storage platform that Proact had previously provided to the university

We believe that even in the long term there will be a mix of internal delivery and purchasing services. There are things that we have to keep in the house for various reasons, including legal requirements.
Hans Jevrell, Enterprise Architect Örebro university

Required to maintain on-premise IT

Örebro University covers most of its IT needs on-premise, but some functions are sourced from service suppliers, including standard cloud services as well as SaaS and PaaS solutions customised for the university.
“We believe that even in the long term there will be a mix of internal delivery and purchasing services. There are things that we have to keep in the house for various reasons, including legal requirements,” says Hans Jevrell.

For its internal IT, the university has consolidated on a small number of products, including for servers and storage. About 80 percent of the servers run in a virtualised environment (VMware). “Because we are a small university, we need to add outside expertise. We can’t maintain every skill in-house, but we also don’t want too many suppliers. Our stringent requirements make it essential to have long-term relationships, with partners who know our environment,” says Hans Jevrell.

One reason why Proact’s solution for the AI server with NetApp storage suited Örebro University so well is that it’s easily scalable. It also blends with the university’s hybrid environment, where it is easy to move data and workloads between our own data center and external SaaS and cloud services. “We are more than happy with the delivery and deployment of the solution, there were no problems at all,” says Hans Jevrell.

Complemented by smaller GPU servers

When the AI computational solution was first deployed, it soon became evident that there was also a need for scaling down, by offering smaller, easily accessible computing resources. The background was that it required a lot of skills to use the large AI machine. This was less of a concern for technically oriented users at the university, who were self-sufficient and knew how to do modeling and organize their datasets to enable AI processing, like machine learning.

But outside the engineering sciences, many researchers needed help for learning to use the platform and how they should apply AI as a research tool. To meet their needs, the university has added smaller servers that offer GPU computing power, using the identical software architecture. This is a comparatively low-cost solution that has proven to work well as a testbed and training environment.

“We wanted to lower the bar for prior knowledge and provide tools and components that make it easier to work with AI. It should be as easy as using a web browser,” says Hans Jevrell.

Facilitating booking and data management

Working groups were started with the task of building an administrative shell around the AI server. This can be likened to a booking office, where users log in and schedule their jobs. They get help managing computational resources, software and algorithms, and not least, moving data in and out of the environment. “It’s like a conference room around the machine. You book it for a certain time and with a specified amount of data, and when admitted, you have it for yourself. Then you just have to clean the ‘room’ when you leave,” says Hans Jevrell.

Any data left in the AI storage area is erased when the next booked project takes over. That’s why there is also separate storage available, where users can save their data models when these are not active. This storage is also protected by regular backups.

“It’s inviting to pay attention to the hardware and the performance it offers, while forgetting about the things you need to use it. But it’s not easy to anticipate everything and we have learned a lot over the course. This has been an enlightening journey for us as well as for our suppliers,” says Hans Jevrell.

Benefits

  • Scalability
    ONTAP AI reference architecture provides a standardised server platform with the option to scale up, or using smaller, compatible GPU-powered servers.
  • Performance
    A large common computational resource, replacing several departmental servers, provides a significant boost in available capacity for AI workloads.
  • Cost effective
    An existing storage system could be upgraded and used as part of the AI computing environment.
  • Security
    Strict requirements on access control, privacy and data protection were met by having the AI environment managed and operated by the University’s IT team.

Key components

  • ONTAP AI reference architecture with NVIDIA GDX-1 computing server and NetApp storage.
  • Small GPU-powered servers for testing and training, complementing the large AI computing resource.
  • In-house developed user portal to facilitate scheduling and data management.
It’s like a conference room around the machine. You book it for a certain time and with a specified amount of data
Hans Jevrell, Enterprise Architect Örebro university

About the customer

Örebro University in Sweden offers a broad range of programs for professional studies, including medicine, psychology, law and engineering. The university has about 15,000 students attending some 100 programs at the undergraduate and graduate level. The university has three faculties and 1,600 employees, of which 400
work in administration, technology and libraries. Örebro University is among the top 500 universities in the world, according to the Times Higher Education 2022 rankings.

www.oru.se

Explore more success stories

#ThePowerOfData

Multicloud backup solution adds to O365 security
NVIDIA DGX Systems creates new opportunities in complex AI-based research
Small footprint & big possibilities thanks to game-changing storage solution

Let’s start something great together!

By clicking Submit, I agree the terms and conditions outlined in the Proact Privacy Policy.

Get in touch

We would love to hear from you. Visit us, call us, join our social media community or send us a message.

By clicking Submit, I agree the terms and conditions outlined in the Proact Privacy Policy.