NCN – Data Centres feature – March 2018 (Page 22/23)
Once the exclusive preserve of major internet and cloud service providers, the benefits of hyperscale-class infrastructure are becoming accessible to colocation business users
Simon Bearne, Commercial Director, Next Generation Data
According to recent research findings from the Cisco Global Cloud Index (2016-2021) hyperscale data centres are on the increase with the relatively small number in existence today expected to climb to over 600 by 2021. With this Cisco predicts these super-sized facilities, also known as mega data centres, will account for 65% of all data deployed globally, handling the vast majority (94 percent) of workloads and compute instances.
Hyperscale facilities were born out of necessity: for supporting the massively scalable architectures of the global web, social media and cloud providers. These have and continue to be at the forefront of driving hyperscale data centre build outs. Hardly a day goes by without an announcement by one or other of them about plans for expanding their regional data centre presence somewhere on the planet. They build them, or in a few cases turn to mega data colocation providers, to seamlessly provision essential compute, memory, network and storage capacity when and wherever needed. In some cases this entails thousands, even millions, of interconnected servers being deployed around the world.
Beyond the hype
Hyperscale data centres are increasingly strategic to the worldwide digital economy. Their rise owes much to the need for much larger and highly secure facilities for hosting exponential growth in web, mobile and public cloud applications and services. They are also essential for powering and cooling the specialist high density environments required for High Performance Computing (HPC), such as those deployed by scientific research and financial services organisations. And Big Data applications where they are needed for managing the vast data storage and analytical requirements created by the Internet of Things and edge computing.
As the nervous system of the digital universe hyperscale facilities have to be highly connected, served by diverse high speed low latency fibre networks capable of receiving and sending large volumes of data around the globe in a matter of milliseconds. By default they are points of presence (PoPs) for leading international carriers and provide access to hundreds of ISPs. Deployment of cloud gateways that bypass the public internet and connect directly into public cloud subsea cable infrastructure, such as Microsoft Azure ExpressRoute, is also a prerequisite.
Furthermore, as they have emerged as the focal points for carriers, ISPs and internet exchanges, hyperscale data centre owners, operators and users are taking more control over the location and specification of global telecom networks. With this shift in the status quo carriers and network providers can no longer assume the leading role. The growing flexibility surrounding location is being helped significantly by the rapidly falling cost of fibre connectivity, allowing the construction of much larger facilities well away from city locations where space and power is often constrained and real estate more expensive.
Data centre power and energy efficiency
Power and connectivity are inextricably linked in hyperscale data centres. While multiple, high speed data networks are their lifeblood, it is power, and plenty of it, which makes high volume data storage and processing possible. The high density racks required to perform such tasks pull considerable amounts of power - some 60 kWs or more – so an abundant and reliable supply is critical. Connection directly to a national grid can significantly boost reliability and ensures smoother transmission.
Hyperscale data centres are also bringing into sharper focus power usage effectiveness (PUE). With cooling typically accounting for 30 – 40 % of a typical data centre’s total energy bill, the more that can be done to optimise and reduce energy consumption and emissions the better; such as using liquid cooling, installing predictive cooling systems; deploying hot and cold aisle containment; more energy efficient servers; and intelligent energy management systems.
Finally, being able to generate most or all power from renewables is a further major requirement in light of increasingly rigorous environmental compliance demanded by governments, campaigners and customers.
On paper, the requirement for more hyperscale facilities offers significant opportunities for colocation providers hoping to gain a slice of the market. However, speaking from our own knowledge and experience, designing, building and operating data centres of this scale and calibre is not for the faint hearted.
Highly specialised construction and electro-mechanical engineering is essential and requires considerable financial investment. Considerable onsite engineering and IT support skills and services are a must too.
Hyperscale customers require their data centres to accommodate different IT infrastructures to the norm such as different types of network switching and cooling; specialised fully automated back-up and auxiliary power systems; and specific power distribution needs for gaining maximum power efficiencies. However, above all, is achieving sufficient economies of scale, ensuring this category of facility remains financially viable as well as affordable to customers.
Fundamentally, these are not just larger and more secure places for housing data compared to smaller conventional data centres. To be able to scale at the velocities required, they must of course offer exceptional space, but at the same time deliver abundant power and deploy high speed network technologies inside and out.
While a tall order there are already a handful of colo operators bringing hyperscale-class facilities within reach of the wider market, whether for colocation purposes or for hosting private and hybrid cloud workloads. This is allowing businesses of all shapes and sizes to benefit from scale economies and vast amounts of expansion space and forwards power; and state of the art infrastructure for optimising resilience, cooling, energy efficiency and network connectivity.