Bubbles before kilowatts
How the innovation hub is being planned. Start with which technologies must sit together and which must stay apart. The power budget follows from the map.
The usual first question in data center planning is how many kilowatts per rack. In an innovation hub, that is the wrong first question.
An innovation hub hosts technologies that are not finished yet. Some are prototypes. Some are deployed elsewhere but have never run alongside each other. Asking each one for a precise power figure produces a stack of guesses. Asking a different question produces a plan.
Start with bubbles
The better question is topological. Treat each technology, or each group of technologies that must run together, as a bubble. Then ask three things about every bubble.
- What has to be isolated? Some workloads need their data kept physically and logically apart from everything else. Some need a room that nobody else can walk into.
- What will eventually merge? Two technologies that run separately today may be designed to become one system later. Those bubbles belong next to each other from day one.
- What does each bubble need besides power? Cooling, water, clean compressed air, dust control, fire suppression, network paths, physical access. Power is one utility among many.
The answer is a map: a set of bubbles, their neighbors, and the walls between them. Only then does it make sense to route power from one corner of a large building to another, and to decide how much redundancy each bubble actually needs.
Redundancy where it matters
Not every bubble needs the same protection. A large building can isolate critical systems into smaller clusters, each with its own backup, instead of wrapping the whole floor in one expensive layer. Smaller clusters are also better for data isolation and security. The difference between protecting everything and protecting the right things is a large share of the capital budget.
Three kinds of technology
The technologies that pass through an innovation hub fall into three groups, and each one is tested differently.
- Inside the data center. Power, cooling, hardware, and the software that runs directly on it. These are measured on the floor itself.
- Inside the enterprise. Software that will live in a customer's environment for security or regulatory reasons, or in a private hosted environment on their behalf. These are tested in isolated zones that mirror how a customer would run them.
- In between. The networks and coordination layers that move data between where it is produced and where it is computed. These are tested across zones, and across distance.
Simulating the real world
A clean lab result means little if it collapses in production. The innovation hub is designed to recreate the conditions that real deployments face: injected latency between racks that stand in for distant sites, coordination across separate zones, and deliberate power disruptions to see what fails and what recovers. A technology that survives those conditions on the floor is far more likely to survive them in the field.
A time axis
Every technology on the floor sits somewhere on a timeline.
- Pre-release. Research and development, where the job is to learn.
- Ready now. Proven enough to build around today.
- Later phases. Important, but not yet. These are planned for, not built for.
Planning a facility means being honest about where each item sits. The facility still needs a power and space budget for all three, but only the middle group drives what gets built first.
Choosing a set
A portfolio of promising technologies is a box of building blocks. They can be assembled many ways, and every combination has different power, space, and security needs. The discipline is choosing which set to build first, writing down the assumptions, and building the muscle to revisit them as the technologies mature. Ambition sets the direction. The floor plan has to be built on what is known today.