A data center, in short, is a dedicated space or building that houses an organization’s IT equipment and servers. The company can draw on its data center resources to operate its business or serve those resources up to the public as a service.
Enterprise data centers and hyperscale data centers can be compared using their scale and performance.
Hyperscale data centers are significantly larger than enterprise data centers, and because of the advantages of economies of scale and custom engineering, they significantly outperform them, too. Not by any means an official definition, a hyperscale data center should exceed 5,000 servers and 10,000 square feet.
What further distinguishes hyperscale data centers is the volume of data, compute, and storage services they process. In a survey, 93% of hyperscale companies expect to have 40 GigaBytes per second (Gbps) or faster network connections. In the same survey, 51% of respondents report that the bandwidth needed to manage vast volumes of data is an increasing challenge.
Comparing their power usage effectiveness (PUE) metrics (although, comparing PUEs between data centers is not always apples to apples,), most enterprise data centers commonly report and average data center PUE between 1.67-1.8. However, Google hyperscale data centers report a PUE of 1.1, where a PUE of 1.0 means perfect efficiency. From a performance standpoint, the comparison is very much like a full-size sedan versus a zero-emissions vehicle.
Hyperscale is also a term that embodies a computing system’s capability to scale, at orders of magnitude, to meet tremendous demand. So hyperscale data centers are exceptionally agile, with the ability to scale up, down, and out to meet any load they service. This can mean adding more compute power, as well as adding more machines, or the ability to scale out to the edge of a network.
Essentially, hyperscale data centers compete “not just [on] scale [alone] but also [on] a certain approach to building and managing infrastructure, emphasizing stripped-down hardware, maximum disaggregation (components can be mixed and matched), modularity, automation, and other principles.”