Bridging the Gap: Transforming the Data Center for the Artificial Intelligence Era

Modern data centers, regardless of size, need to rethink power management and backup strategies, which is a vital part of the innovation strategy.

The era of artificial intelligence will completely change the status quo of data centers. Enterprises of all types are actively exploring how to utilize generative AI technology. This requires them to have more advanced, more secure, and more efficient data center facilities.

Hyperscale is the ideal user group for modern data centers. They have the resources and capabilities to find new opportunities and leverage the most advanced technologies to build new infrastructure.

Still, businesses don’t have to limit themselves to meeting fewer needs. Smaller data centers can be transformed for the AI ​​era by incorporating the right technology to optimize real estate use. This requires careful consideration of aspects such as the computing infrastructure that powers AI applications, new approaches to rack configuration, cooling technologies, and data storage.

This also means strategically considering data center power backup systems to ensure a balanced power strategy for brownfield retrofits. Every piece of data needs backup power, but your existing power equipment is likely taking up space and not adding a penny in revenue. New technology innovations like nickel-zinc (NiZn) batteries provide higher density backup power, potentially increasing backup capacity while freeing up valuable floor space for increased productivity.

Centralized or Distributed Backup Power Supply

To understand the importance of changes in scale, take a look at the following data. McKinsey predicts that data center demand will grow by 10% annually by 2030. By then, demand from the U.S. market alone will reach 35 GW.

The current situation shows that the demand of data center customers has exceeded the carrying capacity of the data center. For large enterprises that are building new data centers or upgrading, increasing density is a solution to provide more computing power per square foot. So it's no surprise that even major cloud service providers are starting to focus on the amount of space taken up by backup power systems.

Normally, data centers are equipped with centralized uninterruptible power supply (UPS) backup systems. In large-scale applications, people are increasingly turning to distributed backup systems, such as server rack battery backup units (BBUs).

Nonprofit organizations such as the Open Compute Project are pushing for new standards using distributed backup power methods. While this approach offers several advantages in hyperscale enterprises, it is not the best choice for colocation facilities or enterprises. This is because the colocation facility would need to accommodate a variety of different tenant configurations, making it impractical to implement. At the same time, for enterprise-class workloads, a decentralized approach may be overkill.

There is also a backup power supply within the server to ensure that the server shuts down normally in the event of a power outage.

These backup systems may or may not complement each other. The key is finding the right mix to ensure power-hungry AI workloads can continue to run. Many modern data center retrofits involve modular infrastructure, giving existing facilities the flexibility to add needed equipment in an iterative manner and with limited space.

Stay away from lead acid

Unfortunately, the lead-acid batteries that have powered data centers for decades are inefficient and take up valuable space. They also have a limited operating temperature range and require more space for cooling technology.

Lead-acid batteries are relatively cheap to start with, but more modern battery technology is worth the investment. Lithium-ion batteries have been on the market for less than a decade, but they have already captured a sizable share of the market in new data center construction. They are more efficient, so take up less floor space, and do not need to be replaced as frequently as lead-acid batteries.

Nickel-zinc battery technology is not as unstable as lead-acid and lithium-ion batteries. In fact, it has no thermal runaway and can operate over a wider temperature range than any competing battery chemistry. Lithium-ion batteries have high energy density, while nickel-zinc batteries have high power density, which means it has a higher power discharge rate. In a backup scenario, when the only goal is to keep the battery running for 15 to 5 minutes or less, you want a small battery that can discharge a lot of power quickly.

Compatibility with legacy equipment

While hyperscalers can start from scratch, enterprises cannot ignore the existing equipment in the data center. Before the introduction of lithium-ion batteries, every data center used lead-acid batteries.

Utilizing the same UPS charging system, data center operators can more easily retrofit nickel-zinc batteries with existing UPS equipment through direct replacement.

At the same time, because lithium batteries require additional protection, it may be easier to replace lead-acid batteries with nickel-zinc batteries than to buy new lithium batteries. Lithium's volatile chemistry creates additional costs in terms of ventilation, high-capacity fire suppression, enhanced indoor combustion ratings, and other safety features not required for nickel-zinc batteries.

The bottom line is that all businesses, regardless of size, need to modernize their data center strategies to keep up with the promise of artificial intelligence. The opportunity to simply build a new data center won't always exist, but the right retrofit strategy will give businesses the impetus for change they need.

The above is the detailed content of Bridging the Gap: Transforming the Data Center for the Artificial Intelligence Era. For more information, please follow other related articles on the PHP Chinese website!