We often talk about utility computing in the cloud, but it can be on-premises too. I like to think of utility from the point of view of the consumer. By concentrating on what the customer experience looks like, we get to avoid pedantic discussions of what is a commodity and what is a cloud. Electricity supply is the usual utility example. Can we get that same experience for computing?
Nobody is surprised that a light comes on when the flick the switch. Nobody thinks about the miles of copper wire and countless switching points that are required to connect the filament in the light bulb to wherever the power is generated. Most people don’t know or care that there is a wholesale market for electricity, with spot pricing and contracted supply. The infrastructure for the utility is largely forgotten. From the point of view of the consumer, it just happens. Much of what cloud services offer is on that same basis. You don’t need to know how it is done, just what will be delivered.
Amazon Web Services (AWS) is the classic example of utility computing. It has an unbelievable amount of complexity that allows millions of consumers to simply access VMs whenever they wish. Of course, the AWS VMs and other services aren’t the end product. Customers want to access applications without thinking about what is behind them. I don’t need to think about the parts that need to work in order to watch a video on Netflix. I just want to search for a movie and settle down to chill. Web services are all about this simple consumption, which hides massive complexity.
This move to focusing on application consumption is important for
enterprise IT, too. IT departments need to return to being enablers for business, rather than being the department of “no” when business comes calling. Spending a lot of time provisioning LUNs and VLANs is not business enablement. Removing the need to spend time on the underlying complexity will enable IT staff to look further up the stack and get closer to the business requirements that should be driving IT. The desire to provide simple consumption for IT infrastructure is central to the rise of hyperconverged systems. These platforms provide a very simple interface that is all about what you do with the infrastructure: run VMs. They hide away every possible implementation and operation detail. Hyperconverged vendors want you to have a simple experience, which they offer by hiding massive complexity.
A central problem for utility computing on-premises is the ability to scale. AWS deploys more infrastructure every week than many enterprise companies’ entire IT estate. But keep in mind that this capacity growth is just ahead of demand growth. Because AWS has such a massive estate, it must deploy a vast amount of equipment each week just to keep up. On the other hand, businesses seldom grow by more than a few percent each year, so the limitation of scale in on-premises may not be as much of a showstopper as some people believe. One of the things that definitely will be limited is the absolute scale. No on-premises deployment will rival AWS for size. This, then, drives the need and ability to automate operations. AWS must operate at massive scale, so everything must be automated, and it has the budget to build the automation that it needs. Most enterprise deployments are much smaller, with much smaller automation budgets. Enterprise customers need prebuilt automation. This is one reason why HCI vendors like Nutanix make a lot of noise about being web-scale. They use technologies that are designed to scale to large numbers of nodes. Other HCI vendors take a different approach. Scale Computing silently automates everything. It wants its infrastructure to disappear for its customers. The Scale platform silently diagnoses and repairs many faults. The Scale support team feeds common support requests to its developers, who then automate the detection and resolution so that customers are never impacted.
We are on a road to simpler consumption of IT. A key starting part is the simplification and automation of the deployment and operation of infrastructure. The more we can forget about our infrastructure and reliably trust that it just works, the better. Spending more time thinking about how to make the business work is much more profitable.