Technical Services

Addressing multiple needs

With a lack of local hyperscale data centers, cloud services, and centralized compute solutions, many enterprises and research institutions in Romania have traditionally exported their digital projects, workloads, and investments outside the country.

In addition to raising data sovereignty and General Data Protection Regulation (GDPR) compliance concerns, these foreign deployments have become exceedingly expensive, due in large part to a sharp rise in electricity prices in the region. To meet the needs of local businesses and address these challenges, ClusterPower set out to build a new, energy-efficient data center in Romania.

ClusterPower also needed to ensure its new data center could provide robust artificial intelligence (AI) application services, such as being able to support training and scaling of next-generation and foundational models, as well as AI inferencing and running real-time applications in a production environment.

Doing so would require a very large-scale AI infrastructure deployment, which has significant compute, energy, and efficiency implications. "Part of our design strategy was to incorporate a flexible, scalable, secure solution into the blueprints of our entire business," said Vladimir Ester, Chief Technology Officer and Co-founder of ClusterPower. "We needed to deploy technical solutions that matched our strategic objectives on our development roadmap."

This meant being able to support the massive compute power that modern AI workloads require, along with the ability to easily scale the compute power as AI workloads evolve and become even more demanding over time. This would require a lot of energy. "We're looking at tens, if not hundreds, of megawatts of IT load that need to be deployed in a very efficient, scalable, and secure environment," said Ester. "We needed to look at solutions that match the outcome of the AI workloads."

To address Romania's need for local, affordable colocation, cloud, and compute services, ClusterPower built a groundbreaking data center campus in Southern Romania that is powered by natural gas and hydrogen and driven by Cisco technologies.

"We have expertise in both IT and energy engineering, and we set out from the very beginning to establish an entirely new, fully sustainable business model," said Ester. "We also wanted our technology foundations and services to be as advanced as our physical infrastructure, and Cisco was the perfect partner."

Sustainability and efficiency

ClusterPower's 25,000-square-meter technology campus is the largest data center facility in Romania and the first in Southeastern Europe with dual Tier III certifications for Design and Facility* from the Uptime Institute. ClusterPower intends to build four similar data centers in other parts of the country in the near future.

Not wanting to rely on—or pay for—public utilities, ClusterPower built its own microgrid that produces energy from gas and hydrogen. Using absorption chillers and combined cooling, heat, and power (CCHP) techniques, the facility captures and recycles heat generated from its high-density technology systems, transforming thermal energy into a cooling agent that is used to air-condition those same systems.

The data center's energy efficiency is enhanced by several Cisco technologies, including Cisco UCS X-Series Modular System consolidated infrastructure to efficiently scale compute power with new and updated server blades. The Cisco Intersight IT operations platform is leveraged to manage the efficiency of the blades, and Cisco Application Centric Infrastructure (Cisco ACI) is utilized to manage thermal conductivity between the blades. "From a sustainability and efficiency perspective, this has a huge impact on how our clients can deploy their workloads and scale their infrastructure in our environment," explained Ester.

"[Together] they are among the most energy-efficient solutions because their modularity allows our cloud platform to orchestrate the workloads in a very efficient manner. This also integrates with our ESG [environmental, social, and governance] objective," said Ester.

"Data centers are some of the largest energy consumers on the planet. At a time when energy resources are limited and costs are rising, it's important to be more sustainable," said Ester. The facility's Power Usage Effectiveness (PUE) rating of 1.1 is roughly 35 percent more efficient than the industry average. "Because we produce our own energy, we save a vast amount of cost and can pass those savings to our customers."

Powerful technology foundation

In addition to pursuing efficiency and sustainability, ClusterPower also wanted a technology foundation that delivers state-of-the-art digital solutions and services, as well as support for the most critical workloads.

"One of our goals is providing highly elastic services that give our customers both flexibility and scalability," Ester explained. "Cisco UCS X is certified for SAP HANA and comes with the latest CPUs and very large memory capacity. It's an extremely powerful platform, and we can scale from a single virtual machine to 10,000 virtual machines in a single day if needed." The total server footprint spans multiple data rooms and is connected to the Cisco ACI multi-pod network fabric.

"It's too tedious and time-consuming to manage tenant environments individually," Ester said. "We're using the multitenancy and microsegmentation features of Cisco ACI to drive operational consistency and efficiency, while maintaining full isolation and protection of each client environment."

Cisco Intersight also aids operational consistency and efficiency across ClusterPower's compute environment. The cloud-based operations platform provides templates and automation that help accelerate infrastructure and service deployments.

"With Cisco Intersight, we have a single platform for managing all of our servers, applications, and client environments," Ester explained. "It's fast and easy to onboard new tenants and customize our services to fit their needs."

Together, this enables ClusterPower to seamlessly upgrade and scale compute power in a live production environment, allowing the company to quickly scale for its customers. "We are now building massive-scale AI infrastructures for training the next generation of foundational models," said Ester. "This is a very challenging project on its own because it has so many implications from both an energy and efficiency perspective for the data center and IT infrastructure."

Replication and expansion

"Having the ability to scale really fast has proven to be a huge advantage for us in this space," explained Ester. "Our underlying foundation is smart enough to adapt to future changes. We've integrated the right technologies that allow us to be flexible, scalable, and more secure for our end customers."

Using its first data center campus as a blueprint, ClusterPower is preparing to build additional facilities and expand its business into new markets. "We are expanding our business, not only within Romania, but also within the region, and constantly improving the blueprint that we already have for the future deployments of data centers," explained Ester.

The Cisco UCS X-Series server architecture will be fully replicated, the data centers will be synchronously connected within the company's Cisco ACI network fabric, and the distributed compute environment will be centrally orchestrated using Cisco Intersight.

"We couldn't do this with any other technology vendor," Ester said. "Cisco has been the best partner to help build something this ambitious, and their computing platforms and network fabric architecture provide the best foundation for client infrastructure and services."

*Uptime Institute Tier Certification overview