NGD’s Steve Davis considers the implications for data centres in the wake of the hyperscale computing phenomenon
There were 320 hyperscale data centres worldwide by the end of 2016 with nearly half in US, according to new Synergy Research data. The US is followed by China and Japan with 20 or so apiece while the UK has 15, more than anywhere else in Europe.
This category of data centre is indeed a rare breed and only a handful of very large global organisations such as the likes of Alibaba, Amazon, Facebook, Google, IBM, Microsoft, Oracle and some major financial institutions currently have a specific need for them; for supporting global internet, Cloud and Big Data services and applications.
But their demand for more hyperscale data centres is growing fast and Synergy expects the number to exceed 400 by end 2018. Clearly meeting the needs of the major public cloud providers and internet/social media leviathans is driving this. However, few independent wholesale or colo operators are capable of building and operating data centres of this scale and calibre, not least because of the massive investment, power and specialist engineering skills required.
Hyperscale user organisations and services providers need hyperscale data centres that can support a single massively scalable architecture which can seamlessly provision and support additional compute, memory, networking and storage capacity as part of a larger computing, distributed computing or grid computing environment. This can mean thousands, even millions of interconnected servers housed in data centres around the world.
Apart from the sheer space and scalable power to rack required for handling large additional increments of compute resources as and when needed, hyperscale users require data centres able to accommodate different IT infrastructures to the norm. Such as servers that are wider than the 19 inch standard, different types of network switching and cooling, and specific power distribution needs for gaining maximum power efficiencies.
At the same time, private cloud and legacy IT infrastructures need accounting for and connecting bearing in mind specific user or government requirements or concerns when it comes to things such as security and data privacy, not to mention utilising existing IT investments.
At NGD we have for some time been aware of the game changing demands that the hyperscale revolution would increasingly place on independent data centre operators. NGD’s world class asset is inherently well suited to hyperscale due to its vast size and all important abundant power supply. The large economies of scale and power available give us the flexibility to quickly and cost-effectively build out custom data halls using blue chip electro-mechanical infrastructure. We have also recruited and trained up a dedicated on-site engineering team capable of installing, configuring and managing hyperscale rack environments as well as cooling.
NGD’s ability to scale as part of a global compute resource infrastructure is also enhanced greatly by already being established as a major point of presence (PoP) with leading international the carriers, our access to several hundred ISPs and deployment of virtual cloud gateways for connecting directly into hyperscale public cloud subsea cable infrastructure - such as Microsoft Azure ExpressRoute.
As the hyperscalers’ thirst for yet more compute resource on a globally interconnected scale continues, it will increasingly become a case of those data centres that can and those that can’t support them.