When You’re HPC Ready…We’re Here for You

According to market research firm Statista, the total amount of data created, captured, copied, and consumed globally is projected to grow to more than 180 zettabytes by 2025. One zettabyte is equal to one trillion gigabytes—a number far beyond our imagination.

With all this data out there, and expanding ways to access the data, more and more enterprises and research institutions are finding they need High Performance Computing (HPC) capabilities to solve advanced computational problems. They’re hiring data scientists to develop the required algorithms and investing capital to purchase enough compute and networking firepower to crunch the numbers as fast as possible.

Perhaps, you’re not yet quite ready for HPC but may be in the future. You may be ready to begin or you’re well into your journey to create the infrastructure and hire the expertise to enable HPC. Either way, consider the data center you will need to provide a home for all the servers, switches, routers, other gear, and the right facility for your environment.

Here’s a quick rundown of what to evaluate on your HPC Journey:

• Density is the first and most important factor—the kilowatts per cabinet your data center can deliver. HPC systems can run from 10 kW per cabinet to as high as 100 kW and above. Your server manufacturer will advise you on the kilowatts your servers require, and using as few cabinets as possible is key. That’s because you want to minimize the length of cables connecting your networking gear. Costs and latency escalate quickly as the distance increases, and your servers may not be able to deliver the performance you require. As an example, if your servers require 100 kW, your networking costs will be a lot lower if you stack all the servers in one 100-kW cabinet rather than using five cabinets that deliver 20 kW each.

• Cooling is another key factor as servers that overheat won’t last long. The higher the kilowatt density of a cabinet, the more heat it generates, so your data center had better have systems to lower the temperature. Depending on your density requirements and your application workload, you may require air cooling, chilled cabinet doors, direct-to-chip cooling, or even chip immersion in a dialectic fluid. Even the physical cabinet layout plays a role with cabinet width and spacing, hot and cold aisle sizing, and air containment all contributing to cooling capabilities.

• Resiliency varies by application according to the mission-critical status. Some applications have geographically diverse nodes that can fail over to each other, reducing the resiliency need. Others are highly dependent on local uptime and require high levels of resiliency. For these applications the data center needs to have sufficient power, backed by Uninterrupted Power Supply systems and Generators, all configured in a redundant topology.

• Cybersecurity is another concern, especially for those organizations subject to regulation. Government data may require FedRAMP compliance while healthcare data will need HIPAA compliance. Other requirements could include SSAE 18, PCI-DSS, and GDPR compliance.

• Location is influenced by whether your personnel need to physically visit the HPC infrastructure. You may also want the infrastructure to be within a certain distance of another node or another provider to reduce latency.

• Equipment Weight increases significantly with High Performance Computing, meaning the flooring of your environment is another concern. While raised floors help diffuse heat, they are limited in how much weight they can support. You may need a data center that offers slab flooring.

Give Yourself Plenty of Time and Look for a Data Center with Flexible Options

One of the biggest pitfalls for enterprises and institutions planning to deploy an HPC environment is underestimating the timeline. Hardware supply chain issues and the time it takes to establish a relationship with a data center provider—plus the time to configure and deploy the gear—usually take several months for an HPC solution. It’s sometimes possible to accelerate the timeline, but we recommend you pick up the phone and call for help at least one year in advance of your target deployment.

No matter where you are on your HPC journey, DataBank is ready to help. Our expertise helps you determine exactly what you need from your data center so that you don’t have to adjust after ordering your servers and networking gear. Our colocation data centers feature a flexible Universal Data Hall Design framework that provides slab floors and allows us to easily add raised flooring to accommodate varying cabinet sizes, layouts, and other options.

Even if you don’t need an HPC solution today, we can provide a cost-effective solution for your current computing needs, and you will have the option to convert to HPC in the future—without moving to another data center.

Interested in learning more about Databank, and how we help our customers with the best high-performance computing options and services to drive their business? Visit www.databank.com today.

For more information on Databank’s HPC solutions, visit our website. We have many years of experience accommodating and delivering high-density colocation HPC solutions in public supercomputing environments such as Georgia Tech and the University of Maryland. We also support publicly traded organizations in the financial services, cloud, software, and AI sectors. Our staff includes seasoned veterans in engineering, construction, and data center operations.