IT loves its buzzwords. But hyperscale is not just a catchy title. It’s the future of technology.
The penny dropped for me around the cloud conversation. It happened a number of years ago, when it was tricky to define cloud, a concept that didn’t seem to fit with the meat-and-potatoes world of IT.
Read this next: Hyperscale on the horizon
I interviewed technology legend Paul Maritz, who had just helped found Pivotal. He raised an example that has since been reused countlessly by many, myself included. Once we had mainframes, which were monolithic systems focused on very specific tasks. Then the client-server PC market rose, allowing for a much wider range of applications. The basic effect of this was that the price per CPU cycle dropped dramatically. As Maritz put it, tracking a mouse on a mainframe was an expensive and wasteful exercise, but on PCs, it was so low-impact that graphic user interfaces became the norm.
This dynamic took another step forward in the ‘90s when thin clients – low-powered machines that channelled the results crunched by large servers – became in vogue, yet not successful. There were several barriers, including the speed of networks and the cost of servers. By the early 2000s, countries such as Japan were lamenting that if nothing was done, datacentres would consume almost all amounts of power capacity generated.
Efficiency became the target for server manufacturers, and at the same time, broadband leapt into prominence. Web browsers were opening doors to operating system-agnostic applications. Hypervisors allowed applications to be virtualised and extend across more than one physical system (in other words, an application was able to view several servers as pooled into one).
Likewise, software-defined concepts started virtualising infrastructure. We didn’t know it yet, but cloud had arrived.
The new IT epoch
The most concise definition for cloud comes from the national Institute of Standards and Technology: "Cloud computing is a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g. networks, servers, storage, applications and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction."
Instead of talking about servers, we could talk about numbers of CPUs or the size of storage pools. workloads – the applications and their functions that run on computers – could be allocated whatever resources they need. Instead of dedicating one server to one application, resources could be thrown at demand on a whim. Amplify that and you have the logical embodiment of hyperscale.
Hyperscale came into its own when companies such as Google started wiring low-powered computers to act together as if they were enormous resource pools that dynamically allocated where demand surfaced. Amazon coined this as ‘elastic compute’, but the common parlance is hyperscale or ‘web-scale’ datacentres/IT.
Just as client-servers drove down costs, these elastic platforms did the same. Ever wondered how Google can provision free Gmail accounts to millions of people or how Facebook shrugs at the billions of content pieces absorbed into it every day? Behind the scenes, hyperscale acts as a titan, shifting its shoulders so the world can get a better deal.
Call it ‘supercomputer on tap’: a million ants can overwhelm an elephant, and a million CPUs can crush even the most daunting of computational demands.
The hyperscale picture
The concept has evolved. Microsoft, Amazon, Google, IBM and other hyperscale providers developed services on the fabric of their hyperscale infrastructure. Amazon’s Echo and Apple’s Siri voice products are effectively services that live on their hyperscale stacks.
“Very few organisations, in 2018, would look at a hyperscale cloud provider as simply a source of virtual machines, networking, and storage,” says Dr Paul Miller, senior analyst at Forrester.
“Microsoft Azure or Amazon Web Services do offer those things, but that’s just table stakes. The real value is in the services the cloud providers offer on top of those basics: things like development tools, rich data management applications, or machine learning, all done at a global scale. For more and more CIOs, their shift to cloud isn’t about finding a new datacentre. It’s about helping – or encouraging – their developers to build applications in a completely different way.”
Many companies are starting to treat hyperscale as a drag racer turning the tap on a nitrous oxide canister, to get that boost when they need it. Got a massive payroll batch job? Throw it at a hyperscale platform.
Newer developments such as containers now make it easier to shift entire applications to where you need them. But hyperscale tools also enable DevOps teams to experiment and exploit new ideas without hefty investments. Development is shifting from multi-month projects to week-long feasibility ‘scrums’.
At the cutting edge, there’s talk about dynamically shifting workloads to the best cloud platform at any given moment. One day you won’t even know where your workload actually is – only that you’re getting the best price/performance ratio within your compliance limits.
Why the ‘digital enterprise’ matters
This is advanced stuff, which is why the whole ‘digital enterprise’ narrative is so urgent. Big wave surfers need a jet ski to reach the speeds of the watery behemoths they mount. Likewise, companies can’t expect to paddle into hyperscale and do anything else but drown.
Hyperscale isn’t the only game in town. Cloud models can be blended and there are even ‘on-premise’ flavours such as Azure Stack. Nonetheless, companies are leveraging hyperscale’s power and costs to their advantage. Likewise, hyperscale providers are investing big: $75 billion in 2017, according to the Synergy Group, with Google, Microsoft, Amazon, Apple and Facebook leading the pack. It’s a global arms race: Chinese giants Alibaba and Tencent nip at the heels of their Western peers.
There’s growing interest in replicating hyperscale or webscale in smaller capacities. This is promoted by concepts such as edge computing (taking some of cloud’s compute and storage capabilities on-site to reduce latency), supported by hyper-converged systems vendors á la Dell EMC and HP, and datacentre infrastructurein-a-box promoted by the likes of Schneider Electric. Hundreds of hyperscale datacentres already exist, with more being added every month. Africa used to be left out, but the imminent arrival of Azure in South Africa, as well as the expanding footprints of Amazon web Services and Google, is changing this – and they're being driven by market demand.
Hyperscale is beyond IT hype. It’s the new golden path that brings digital into our lives. from smart cities to home security to the ERP to your favourite app: somewhere behind the scenes, hyperscale is proving the torque to keep those wheels spinning.