The modern data centre is an intricate balance of components. IT equipment interacts with power and cooling infrastructure in complex ways, and each facility has its own unique cost and energy efficiency profile.
The ability to accurately predict what that profile will be requires some complex mathematics; relevant variables include the precise hardware configuration, ambient temperatures and fluctuations in IT load, to name but a few.
Engineering services provider Arup designs and develops data centre facilities for clients including hosting providers and financial services companies. Modelling these variables mathematically allows it to calculate the ongoing impact that design choices will have on both the cost and energy consumption of its clients’ facilities.
This scientific approach to data centre management is sorely needed, says Andrew Harrison, who leads Arup’s global science and industry business. Harrison argues that most data centre management tools do not match the sophistication of the facilities themselves.
“One of our customers uses the analogy of car maintenance: their data centre is like a Formula 1 car, and they only have the tools to manage a Ford,” he says.
Appliance of Science – Data centre modelling tool Prognose
Traditionally, Arup’s engineers have used spreadsheet software to model data centre performance. “These spreadsheets can be quite complicated, but that’s fine if you’re the person who built them,” he explains.
But there are a number of drawbacks to the spreadsheet approach, he says. Firstly, it is not easy to ensure that data is used consistently across the organisation. And secondly, these complex spreadsheets can be difficult for clients to understand, especially if they do not have an engineering background themselves.
Since 2009, Arup has used a data centre modelling tool named Prognose from IT management software supplier Romonet. Built on a database of data centre equipment, Prognose allows the user to model various combinations and configurations, and to calculate the resulting energy efficiency and cost profile.
Importantly, Prognose presents the results of these calculations in a graphical format, making them accessible to clients. “They don’t want big reports, they want nice clear graphical illustration of the benefits,” explains Harrison.
And while Arup’s engineers were able to do the calculations previously, Prognose allows them to do them faster. This means that they can spend more time exploring hypothetical configurations, and are therefore more likely to find the perfect balance. “In the typical data centre design process, people define the problem they want to solve and focus all their attention on that,” Harrison explains. “Prognose allows us to try out different scenarios and see what the outcome might be.”
An example of this comes from a data centre project that Arup recently conducted with a client. “We were trying to minimise the capital expenditure required for the build,” Harrison recalls. “By mapping out the projected IT loads, we were able to demonstrate that the client could delay building one of the generators until the third year of the project.”
In another project, Arup helped a financial services company calculate the impact that using in-row cooling, rather than air conditioning, would have on the ratio of operational to capital expenditure.
Harrison acknowledges that most companies are not involved in multiple data centre builds every year, and so might not get the value from Prognose that Arup has done.
But he adds that data centre operators of all sizes would do well to take a more analytical approach to the ongoing management of their facilities. The way in which data centre components interact is often examined during the design phase, he says, but it is usually a “one off”. That means that the impact of introducing new equipment further down the line cannot be fully understood.
“We advocate that our clients take an MOT approach,” says Harrison. “An annual audit helps you make sure that you are clear about what you’ve got, and what your operating procedures are.”