If your data center is reaching capacity and you’re thinking about cracking open the corporate piggy bank to fund a new data center, stop right there.
By following some simple best practices, you may be able to take your existing data center and retrofit it to last pretty much forever, says Gartner analyst David Cappuccio.
“If you do it right, there’s a good chance you could live in a fairly well designed data center for decades,” Cappuccio says.
So, how do you get there? First, you need to identify the goals of the infinite data center. It needs to be energy efficient. It needs to be economical to build. It needs to be able to adapt to new technologies. And it needs to be able to support continuous growth.
The first step is to build up the density of your server racks. If your racks are at 50% or 60% capacity, then consider increasing the density to 80% or even 90%, says Cappuccio.
Increasing the density allows you to run your existing workloads on a smaller footprint and to create space for future growth.
The reasons data center managers don’t run their server racks at 90% capacity relates mostly to the concern that the servers will run too hot.
Cappuccio recommends that data center managers analyze the different types of activities running on those servers and to create high density, medium density and low density zones.
Throwing up walls and creating a separate room for high density servers allows you to deliver the requisite cooling to that room, but then allows you to run your storage devices and telecom gear, for example, in a zone that doesn’t require state-of-the-art cooling systems.
Creating these speciality zones can save companies up to 40% in operating expenses, Cappuccio says.
The next step is to embark on a slow, systematic refresh cycle. Buy a new rack of state-of-the-art 1U servers and methodically move workloads onto this new rack until it’s up to 80% or 90% capacity.
Cappuccio argues that the energy efficiency of the new server will pay for itself over time, compared to running older, energy inefficient servers. To put a number on it, Cappuccio estimates that the energy savings can fund 80% of refresh costs.
And as you move workloads to the new server, you can decommission older server racks and thereby free up additional space. This becomes an ongoing process that allows you to continue to improve the efficiency of your existing data center without ever having to construct a new building.
There are other tips and tricks for building data center efficiency, Cappuccio says. He recommends keeping the data center at 78 degrees. If you’re currently running your data center at 72 degrees, Cappuccio points out that you save 2% to 3% in cooling costs for each degree that you raise the average temperature in the data center.
He also says that there are better ways to approach the cooling issue. Many data centers today have hot aisles and cold aisles. This requires an extensive system of cold air being forced up through the floor and pushed through the servers via fans. Then the hot air comes out the back and is sucked up into the ceiling, where it is cooled and circulated through the loop.
Cappuccio says it makes more sense to build a smaller, self-contained high-density room within the data center for the racks that need it, and then to have lower density zones that don’t require the same level of cooling.
He also says there are new, liquid cooling technologies that could help cool servers more efficiently. These liquid cooling techniques, which include putting refrigerator coils on the server door itself, are much more expensive, but pay for themselves in the long run.
And the final component to the infinite data center plan is the cloud. Cappuccio says companies should think about moving non-critical workloads, like testing or archiving, to a cloud provider.
Eventually, companies will create a hybrid cloud strategy where certain workloads live in the cloud, thus freeing up space in the immortal data center.