Data centres at scale: just for the behemoths?

The key to the cloud services which meet customer is hyperscale which is designed and operated to meet the ‘everything now’ expectations created by digital technology.

What is ‘scale’?

‘Scale’ as a principle of data centre design and operation represents an inter-related series of attributes – the capability to match IT resources to meet demand (scaling up/scaling down) based on the standardization of datacenter components to create the integrated and systematized environment where this is possible. In real terms, scalability represents a new era of design and operation for data centres although it represents an established operational objective.

Why has scale become important?

Nick ParfittThere are many reasons for the emergence of scale as a primary operational objective. Much has been written about the increasing volumes of data generated both by humans and especially by machines. In consequence, businesses are collecting thousands of data points on customers today and data centres are processing more information than ever before. More people have access to the Internet via mobile devices and internet and social media is collecting data of value to analysts and advertisers from billions of people. Increased computing capacity is adding to the burgeoning pool of data in the cloud. As businesses make the transition towards a truly digital ecosystem, breaking down geographical barriers and entering new markets, the need for storing, analyzing and using this ocean of data is rising. Understanding this need, it is crucial for data centers to scale up as their load becomes greater and more variable and as the demand for apps and services that are instantaneous, high quality and always available becomes the norm.

The key to the cloud services which meet customer is hyperscale which is designed and operated to meet the ‘everything now’ expectations created by digital technology. The development of hyperscale and webscale have very much arrived in India driven by the nation’s status as one of the world’s key growth markets for consumer and business demand for IT hardware and services. With 451 million active internet users in India, the volume of data generated is considerable, growing at an estimated CAGR of 25%. Businesses are building scalable platforms relying on data centers to provide the infrastructure to store and process this data. and the development of network infrastructure and the supply of power, water and skilled staff has reached a point where hyperscale is a viable option.

What can be learned from hyperscale

In DCD’s research of January 2018, over 80% of 156 respondents from India agreed that smaller data centres could learn from the major cloud and colocation providers. The situation that the ‘giants’ face in terms of dealing with managing capacity to meet demand is different in scale but not necessarily in kind from most Indian businesses which have any IT presence. It indicates the key role of analytics in identifying demand variability in order to inform both data centre capacity planning and, more broadly, corporate digital infrastructure strategy. The hyperscale data centre requires a considerable amount of power and the impact of efficiency on the cloud provider bottom line will therefore be a considerable focus –this has led to initiatives such as customized hardware. While most smaller data centres will not be gain much advantage from such initiatives, the focus of data centre giants on management software, automation, and increasingly on AI-driven operation as well. This is a path that smaller operators can follow and are doing so. The cloud giants are conscious of the direct contribution that the performance of their data centres make to their brand and as data centres move towards greater public awareness for both positive and negative reasons, so this is a perspective that enterprise would do well to follow.

[This is an authored article by Nick Parfitt, Senior Global Research Analyst, DCD Group. All views, opinions and expressions are personal and limited to the author.]