Before you buy a new house, you measure its cost against its potential future value. You assess utility bills and calculate whether you could afford a yearly increase. You even decide what furniture you need or whether there are better-styled alternatives to existing items.
This is essentially the same process a business should take when selecting a site for a new data center. How much does land cost? What about the unrecoverable costs of construction and are they all manageable? Is there renewable power available? Will hardware from one vendor meet environmental and efficiency requirements over another supplier?
These assessments and simulations sound simple to execute, although often these calculations are conducted with paper mathematics, error-prone spreadsheets and human intuition. This is imprudent when you consider hundreds of millions of dollars are at stake.
If the business is to ensure informed decision-making, accurately predicting data center operational performance, calculating prospective capacity and energy consumption, and determining the likely total cost of ownership is critical.
Leaving Nothing to Chance
Predictive analytics business applications are increasingly being deployed to model facilities before a single dollar is spent.
Capital investment is the main driver behind this – i.e. how much will everything cost and how could we reduce the CAPEX budget? Data centers are not ‘on-a-whim’ decisions – they are major corporate financing initiatives. If delivered over-budget, they have massive negative implications on a company’s balance sheet and private investment potential.
Land costs are a typical example within this context. It is common practice to sell land for data center development at a premium, much higher than if it was used for agricultural purposes. What has to be factored in, however, is what else that location offers. A case in point is Iowa, where Microsoft, Facebook and Google have all paid ‘above market rates’ for land.
In each case the overall cost of the projects dwarfed the land costs but more importantly each organization took advantage of a plentiful supply of wind power and zero tax on electricity costs for large projects. Iowa also offers superior fiber cabling, which given the importance of connectivity and low latency data transmittance, ticks another crucial box.
Capacity is equally important when planning a data center as downtime must be avoided at all costs. Every square foot of space should be used and if the design of the data center cannot handle a maximum workload, then the investment will not reach its full potential. This is the type of information a business needs to confirm before building a data center, especially if the facility is a colocation facility where availability is paramount to business success.
All of these inter-related factors must be included within a model to ensure capital expenditure is wisely allocated.
The Wider Value of Analytics
When it comes to environmental influences, it is no longer enough to just consider long-term power consumption and the monthly recurring energy bill. Prior to construction, impact assessments have to be made using a wide range of metrics.
Google believed that it was boosting its green credentials when it decided to use wind energy at a data center in Oklahoma. However, questions were asked about what else that energy could be used for instead. These considerations must be part of the impact assessment.
Water availability and conservation are also high on the environmental agenda. The Phoenix Metro area in Arizona is a popular location for data centers, but not suitable if the cooling of server rooms relies on water, something the area has a distinct lack of. If a data center uses too much water, there are substantial CSR risks. After all, Greenpeace is not a forgiving organization.
Some companies take advantage of naturally cooler climates and locate data centers underground or close to mountains and rivers. This helps to reduce cooling costs and provides a natural, plentiful source of energy, but do the financial savings over, say, a 20-year period outweigh the cost of traditional mechanical systems?
The Green Mountain data center, built into the side of a Norwegian mountain, retains a deep pool of consistently cold water which the facility uses for cooling, but such an innovative facility can hardly be moved once construction is finished, can it? Prior to the build stage an organization must be certain the majority of unrecoverable costs have been accounted for.
The Devil is in the Detail
When it comes to predicting future operating expenditure, energy is likely to come at the top of the list, and site location can also make a significant difference.
Rates in California are high in comparison with other states, including Oregon, where there are tax free enterprise zones, but this can also be negated by the existing availability of energy to a site. Tax on energy should be part of the calculations for a facility, even to the extent of looking at incentives being offered overseas. Policy changes in Nordic countries are allowing large-scale data centers to take advantage of very low energy tax rates.
Undoubtedly organizations give attention to these different considerations, but to achieve an accurate, accountable evaluation entails analysis of multiple scenarios and the ability to pinpoint predicted costs and performance levels.
This can only be achieved by using applications that model, simulate and predict data center investment and operations. Importantly, this means analysis covering the entire development of the project that takes into account possible changes in energy costs, decreases in taxes and other longer term eventualities.
Richard Jenkins, Global Marketing & Strategic Partnerships, Romonet
Internationally-focused sales and marketing professional in enterprise software, data center analytics, IoT, cleantech and multimedia. On leaving the Royal Navy in 1990, Richard worked for UK-based IT resellers developing channels across EMEA, building and selling a Y2K IT contractor firm, and providing IT consulting services to London-based banks. Other positions include Tivoli/IBM and Crystal Decisions (later Business Objects), Corporate Radar, Kyoto Planet, Plantiga and RF Code.
Subscribe to Data Informed for the latest information and news on big data and analytics for the enterprise.