Can data centers built long before GenAI support new workloads? That’s one of the questions I hear at industry conferences and when I’m on technology panels lately, as more organizations integrate AI into their applications and infrastructure. My apology for beginning this blog with a spoiler, but the answer is a resounding “Yes!” Read on to learn why.
There are IT professionals and business executives who are unsure whether data centers built in the 1980s, 90s and even the early 2000s can meet the power, environmental and interconnectivity demands of high-performance computing required by next-generation applications.
I’m happy to say that in most instances I can tell clients and prospects that CoreSite’s data centers can support their requirements as they integrate newer, more demanding applications into their business.
That’s because most of CoreSite’s data centers were designed and architected with future performance requirements in mind. For example, our data centers were incorporated with redundant switch gear, as well as spare generator positions, spare electrical distribution positions and spare pumps. The buildings were built with formidable structural integrity in anticipation of heavier servers and power and cooling equipment.
As for power requirements, while most older colocation facilities were designed to support five-to-eight kilowatts per cabinet, CoreSite data centers always have been able to support upwards of 40 kilowatts per cabinet with air cooling. Liquid cooling is a hot topic, but the fact is liquid cooling has been in use for decades. Most of CoreSite’s older data centers have piping that enables adding taps for liquid cooling.
Data center retrofitting is the process of upgrading, improving and adding new equipment to meet customers’ evolving business requirements, improve efficiency, and to squeeze more capabilities out of existing facilities without sacrificing performance, flexibility or reliability.
AI, ML and other high-performance applications place new cooling demands on data center operators. Prior to the advent of AI, average servers in data centers generated about 1.5 kilowatts of heat. However, as my colleague, Anthony White, discusses in his blog, Solving Tomorrow’s Cooling Challenges with Liquid Cooling, “The amount of power used by a fully populated server rack, measured in kilowatts per rack is rising, and higher power density generates more heat.” Average server rack density has increased to 12 kW, according to a recent AFCOM survey, and should only continue to increase.
Another colleague, Joseph Liccardo, elaborates on this in his blog Data Center Cooling Challenges, saying, “Because liquid is more efficient than air in transferring heat, it can be more effective than air cooling. It also can support greater server densities and high-power chips that generate a lot of heat.”
Before AI can begin making good decisions for its users, AI models must be fed enormous volumes of data. CoreSite data centers are hubs for interconnection, providing broad and efficient access to the data sources that train AI models. CoreSite’s network-dense data centers can serve as funnels for data produced at every end point, by every connected device.
CoreSite data centers were also designed to meet increasing Internet Service Provider connectivity and enterprise-to-enterprise interconnection requirements. Many of our data centers first were carrier hotels, with extensive network interconnection capabilities. Existing CoreSite data centers are already being used by public and private cloud providers, as well as carriers, as hubs and network POPs, strongly positioning them for new bandwidth demands. Today, CoreSite data centers offer direct connections to all major cloud providers, providing low-latency access to cloud services in addition to excellent support for hybrid IT.
This prescience regarding extremely robust, low latency interconnection has proved especially important, given the insatiable data feeding needs of AI and ML applications and overall increase in broadband usage created by HD streaming, telemedicine, remote work, advanced financial services, etc.
No one could have foreseen the explosion of high-density computing inherent in AI, ML and other applications that require ever-more specialized and robust infrastructure. The good news is that organizations don’t have to wait anywhere from two-to-seven years for new data centers to be built and become operational to put these rapidly evolving technologies to work.
By creatively, intentionally and opportunistically retrofitting existing data centers, CoreSite has been able to meet the immediate requirements of its customers and prospects, enabling them to leverage these new applications to remain competitive. We look for opportunities during our maintenance and end-of-life replacement activities to make improvements that will future-enable our facilities and do not disrupt the IT operations of our current customers.
At CoreSite, our intent is to operate data centers that are designed and maintained in a way that enables us to meet the needs of all types of organizations – enabling support for high-, medium- and low-density workloads within the same facility or campus.
As I said at the beginning of this blog, it’s not necessary for organizations to put their next-generation strategies and projects on hold and wait for new data centers to be built and opened.
Well-maintained and retrofitted facilities, like those CoreSite operates, are still very well-positioned to support the needs of all types of organizations. They can and do support the full range of workloads and deliver exceptional performance and flexibility to customers.
At the same time, CoreSite is growing its data center portfolio through the expansion of existing campuses, the acquisition of existing state-of-the-art data centers and through new data center construction driven by customer demand and requirements.
Know More
CoreSite data centers are designed to be AI-enablers, where customers can develop and execute long-term high density workload implementation strategies. CoreSite is certified by the NVIDIA DGXTM-Ready Data Center Program, and offers the power density, energy-efficient designs, liquid cooling solutions and robust network interconnections needed for AI and other high-performance workloads.
Download the position paper, "Trust CoreSite Data Centers to Enable Your AI Strategy" to learn more about how CoreSite colocation data centers can help transform your business into a future-ready enterprise.