The future of data centers isn't bigger; it's smaller and adaptable, with workloads consolidated, dynamic and distributed elsewhere. Here's how a data center layout should look for that paradigm.
The world of IT is changing; application rationalization, hardware virtualization and consolidation leave organizations with large legacy facilities housing 50% or less of what they were previously running. New, high-density server, storage and network equipment -- along with converged infrastructure systems such as VCE Vblock Systems, Cisco Unified Data Center, IBM PureSystems and Dell Active Systems -- mean that less space is required for more effective direct business compute power.
With the introduction of cloud computing, data center managers and systems architects no longer just have to decide how best to support a workload. Now, they also have to decide through what means they'll do it. A workload that would normally sit on a stack totally owned by the organization might now live at a colocation facility or be outsourced through infrastructure, platform or software as a service (IaaS, PaaS or SaaS).
Even workloads kept in-house might not always stay there, so it doesn't make sense to design a data center to run certain workloads interminably. Cloud is on the cusp of being a mainstream platform, and organizations that have built a data center for hosting a specific application over a long period of time could be at a disadvantage.
Flexible data center layout
When you're designing a data center for the future, consider the facility and the IT hardware that it supports. From a hardware point of view, a full IT lifecycle management approach maintains a dynamic infrastructure. Hardware assets can grow and shrink as needs change, with excess equipment sold off to recoup some costs. With subscription pricing, IT organizations control software licenses better, signing up or shutting down subscriptions as required.
The main issues, however, revolve around the facility. A data center is a mostly fixed asset -- if it is 10,000 square feet and the business finds that it only needs 5,000 square feet to house the IT workload, the walls cannot converge. Even if half the space can be repurposed -- for example, by walling off office space -- that solves only a small part of the problem.
A legacy data center facility has a fixed layout for services. Power distribution units (PDUs) and computer room air conditioner (CRAC) systems are fixed in place; uninterruptible power systems (UPS) and auxiliary generators are sized to suit the original data center design. Consider the following when mapping your future data center layout:
Floor plan. The first place to start building tomorrow's data center is with the physical design. If sloping subfloors and raised data-center floors are preferred to deal with flooding risks, then create multiple gullies rather than a single, V-shaped system. This way, if downsizing is required, raised walls will mark off each gulley that can be used to build new walls without affecting the capabilities of the subfloor to allow drainage for the data center itself.
Cabling. Data center cabling needs to be fully structured. Carry data and power through separate paths, and build in an easy means of re-laying cables, should the facility layout change.
Power. Rather than build PDUs against walls or pillars, make them freestanding, with power feeds coming from structured cabling at the roof. In a redesign, power distribution is relocated as flexibly as the rest of the IT equipment.
For UPSes and auxiliary generators, a monolithic system requires that the organization either massively over-engineer the system for all contingencies or buy a new unit as data center needs change. As most UPSes are in-line, every efficiency percentage loss could be against the rating of the UPS -- not against the actual power consumed by the IT equipment.
Generators' fuel usage is pretty much in line with the product's rating, so even when a generator is running below its rated power, it will use a lot more fuel than an appropriately engineered setup.
Cooling. A move to free-air cooling or other low-energy system will reduce the impact of redesigning cooling when the data center changes size. If combined with effective hot and cold aisle approaches with ducted cooling, the cooling system will be appropriate to the need, with less worry about placement.
When the data center design calls for a CRAC-based system, choose a more modular system with multiple, balanced, variable-speed CRAC units.