It's been a truism from the dawn of cloud computing (and its predecessors) that the underlying hardware is irrelevant to the end user (ie the enterprise). And as a truism, it holds up pretty well: of course, applications are paramount. However, as is so often the case, it is not always quite as simple as that: in fact, you do care about the hardware.
Enterprise cloud provision is dying
The dawn of enterprise cloud computing began with a raft of companies pitching for business by positioning themselves as enterprise ready, which meant that they understood enterprises' needs. Cloud providers in this category included very well-known names - you probably know who they were.
This kind of positioning was a bid to separate themselves from those companies offering consumer-level services, and was driven by the need - and appropriately priced - to amortise the costs of the top-end, enterprise-grade hardware in which they had invested. High levels of expenditure on super-reliable hardware would deliver very high reliability levels, was the reasoning, and would be attractive to enterprises who would come to rely on such provision.
However, the notion of such specialised, enterprise cloud provision is dying. A number of large providers are no longer offering cloud services for the simple reason that not enough enterprises have taken up the offer. They found that cloud providers delivering services using commodity hardware are still able to offer services with high levels of reliability, and that what matters more is both the range of solutions they are able to offer and the price.
Commodity rules OK
Modern cloud service providers rely on large pools of commodity hardware to create a utility computing environment - an approach dubbed hyper-scale. This allows them to use virtualised storage, network and compute resources that are deployed on demand by provisioning policies set at deployment time.
And how does this affect reliability? Instead of expecting hardware to never or very rarely fail, hardware failures are anticipated, and managed so that they have minimal or zero impact on customer services. In other words, this approach does not focus the provider's expenditure on reaching for the last few percentage points of reliability. Instead, buying lower-cost equipment and packing systems densely results in economies of scale due to reduced space, cooling and energy requirements. Those savings can then be passed on to the customer.
In this way, the large cloud provider can offer reliability as good if not better than those deploying more expensive kit, while gaining greater control over specifications, adding features that are required in a multi-tenanted, hyper-scale cloud environment, and leaving out those that are not.
Does hardware matter?
So the question remains: do you care about the hardware when buying cloud services? When you consider that a cloud provider offers services, solutions and applications, the answer has to be, on the face of it, no. But when when you look at those cloud providers who offer the richest sets of enterprise services, it is clear that such offerings are underpinned by hyper-scale hardware.
The hardware is not the responsibility of enterprise buyers but it is delivering applications reliably at a price which enterprises are prepared to pay. Which makes the hardware very important indeed.