Hyperscale Data Centers
Size: Regular data centers are typically smaller, usually less than 2000 square meters, and house hundreds of racks. In contrast, hyperscale data centers are massive, often exceeding 10,000 square meters and containing thousands of racks. The size difference reflects the scale of operations each type of data center is designed to handle.
Power Substation: Regular data centers typically have a medium voltage (MV) substation, with either a single or two power feeds. Hyperscale data centers, however, have an onsite high voltage/medium voltage (HV/MV) substation with N+1 redundancy. This ensures a more reliable power supply, which is crucial for maintaining the high availability required in hyperscale data centers.
Scalability: Is a key feature of any data center. Regular data centers typically double their initial capacity in Day 2 expansion, using a Load-on-Demand approach. Hyperscale data centers, however, scale in multiples of 4, 5, 6, 8, 10, or 12 MW, with fast incremental deployments until reaching maximum power capacity. This allows them to quickly scale up operations to meet increasing demand.
Modularity: Regular data centers typically use a Shell & Core approach in Day 1, with Fit-to-finish for future expansion. Hyperscale data centers use a Master Planned Phased Construction approach, which allows for more efficient use of space and resources.
Efficiency: Is another key consideration in data center design. Regular data centers typically have lower utilization and less efficient resource allocation, with N+N redundancy. Hyperscale data centers, on the other hand, have high utilization and optimized resource allocation, using N+1 Distributed Resiliency for up to 6 to 1 or 6/5 redundancy. This results in more efficient use of resources and lower operating costs.
Resiliency: Is crucial in data center operations. Regular data centers are reliant on manual intervention, leading to a higher risk of downtime due to unplanned events. They typically offer four 9s’ availability. Hyperscale data centers, however, are fault-tolerant and concurrently maintainable, with automatic failover. They offer up to five 9s’ availability, reducing the risk of downtime and ensuring a more reliable service.
Technology: Regular data centers typically use standard systems and components, with limited technology options. Hyperscale data centers, however, use cutting-edge technology and offer customized/bespoke solutions. This allows them to better meet the specific needs of their clients and stay at the forefront of technological advancements.
Connectivity: Is another key feature of data centers. Regular data centers typically have two fiber entries and either a single or two internet service providers. Hyperscale data centers, however, offer multiple degrees of fiber, are carrier-neutral, and offer Metro Ethernet Networks and high-speed 400-800G connectivity. This ensures a more reliable and faster connection, which is crucial for the large-scale operations they support.
CAPEX & OPEX Cost: In terms of capital and operational costs, regular data centers typically have higher costs per MW. Hyperscale data centers, however, have lower capital and operational costs per MW, due to economies of scale. This makes them a more cost-effective option for large-scale operations.
Target Customers: Regular data centers typically target businesses of all sizes, government organizations, and banking & financial institutions. Hyperscale data centers, on the other hand, target large enterprises or government agencies, cloud service providers, and tech giants. This reflects the different scales of operations each type of data center is designed to support.
Focus: Regular data centers typically focus on a variety of needs, often prioritizing cost and reliability. Hyperscale data centers, however, focus on speed, efficiency, high availability, and large-scale data processing. This reflects the different priorities of their target customers.
Common Workloads: Regular data centers typically host line of business applications, hosting, databases, and storage. Hyperscale data centers, however, support cloud services, large-scale data processing, content delivery networks (CDN), and artificial intelligence/machine learning (AI/ML) workloads. This reflects the different types of workloads each type of data center is designed to handle.
Sustainability: Regular data centers typically have voluntary reporting, are energy efficient, with a Power Usage Effectiveness (PUE) between 1.5 and 2, and have moderate Environmental, Social, and Governance (ESG) goals. Hyperscale data centers, however, have mandatory reporting, low PUE (<1.4), low Water Usage Effectiveness (WUE) and Carbon Usage Effectiveness (CUE), use renewable energy and energy reuse, and are carbon neutral. This reflects the increasing importance of sustainability in data center operations.
In conclusion, while regular and hyperscale data centers share many similarities, they also have key differences that reflect their different scales of operations and target customers. Understanding these differences is crucial for choosing the right type of data center for your needs.