Hyperconverged Infrastructure

Hyperconverged Infrastructure Is Evergreen


There is a Bizzaro cartoon from the early 90s that says "actually, we prefer the 'sparse look' – we just can't afford a place big enough to achieve it." One the face of it, this is simply an amusing single panel cartoon with a biting critique of western materialism. Underneath it, however, is a layered discussion about human nature that has real-world implications for organizations trying to manage the lifecycle of their data centers.

When it comes to home décor, minimalism is prized. We collectively look down on clutter, and in our media the living spaces of the well-to-do are always replete with wide open spaces, and generally free of knick-knacks. In the real world, humans accumulate material objects easily, and aren't very good at giving them up.

We like simplicity. Our brains have to expend effort to filter out extraneous sensory input, even when that sensory input is visual. Clutter impacts some individuals more than others, but evidence suggests everyone is impacted by ego depletion.

Ego depletion is a hypothesis that states that we all wake up in the morning with a limited amount of mental resources available to be expended. Once expended, our ability to exert willpower – and in some cases, even make rational decisions – is compromised. While the details of the ego depletion hypothesis are still begin debated, the basic concept – that precious few of us have unlimited willpower or other cognitive resources – seems to reflect most people's experiences, and explains many popular aesthetic sensibilities.

Ego depletion has a direct parallel in IT. Decision fatigue is a very real concern for all individuals and organizations. IT staffs are particularly susceptible to it: IT consists of innumerable products from a multitude of vendors, all of which are constantly changing and evolving. IT staffs are constantly asked to learn new management interfaces, and keep track of dozens of infrastructure elements from potentially hundreds of vendors. It can all be a little overwhelming.

Another parallel is that humans don't like to throw out anything that they feel might have a future use, even if there’s no immediate use in the present. In a home, this leads to clutter. In a data center, it leads to "organic" or "brownfield" data centers, where multiple generations of technology are made to coexist, and infrastructure from previous refreshes is repurposed as new equipment is purchased.

Like the home of the lady in the Bizarro cartoon, our data centers have limited space. It may save on capital expenditure to demote last cycle's primary workload infrastructure to secondary workload duties, but doing so poses a few problems.

Equipment from the last refresh cycle isn't capable of the density or workload consolidation that the modern infrastructure can deliver, requiring more data center space than if the organization had done a full refresh. In addition, staged obsolescence can magnify the proliferation of management interfaces discussed above, and usually comes with disruptions at refresh time as workloads are transitioned from one generation's infrastructure to the next.

The Art of Minimalism

Hyperconvergence can be a solution to several of the above problems. By combining storage, compute, and networking, hyperconvergence reduces the management overhead for administrators while delivering basic IT automation by integrating the three basic elements of a virtual infrastructure.

In addition, hyperconverged clusters are designed to allow multiple generations of nodes to coexist in a single cluster. This allows both capacity and performance to be added as needed. More importantly, it allows nodes to be retired when they’re no longer economical to operate, with both addition and removal of nodes from a cluster being a non-disruptive process.

Properly planned, hyperconverged infrastructure (HCI) allows organizations to maintain an "always evergreen" approach to IT infrastructure, phasing in new functionality in a non-disruptive fashion, and phasing out older equipment in the same way.

Treating individual nodes as completely interchangeable, disposable commodities changes the psychology of IT refreshes. If the only real difference between the last refresh’s solution and this refresh's solution are easily quantifiable metrics like performance or workload density, then the natural tendency to hoard older equipment is diminished.

New HCI nodes don't bring a new management interface. Their value isn't subjective, and there’s no distinction between the infrastructure that primary and secondary workloads operate on.

Data centers with no space concerns can operate nodes until they reach end of support, and then discard them. Data centers tight on space will make economic judgements based on workload density. Either way, HCI provides the means to achieve the "sparse look," if only because it helps take the worst tendencies of the human decision makers out of the equation.