Data Center Forecasting: Reducing Risks and Over Provisioning

WiredRE Data Center Forecasting Q4 2014In the modern data center industry, IT budgets of between $50 and $100 million are being allocated by self-taught IT staff unfamiliar with the current state of the market. WiredRE research shows that an average data center is over provisioned by between 25 and 50 percent and nine out of 10 facilities in the U.S. are over provisioned, making it the number one risk in cloud and data center planning.

The data center construction market is currently worth $14 billion and is projected to grow to $22 billion in the next five years. The total estimated spending on IT increased 3.2 percent from 2013, reaching $3.8 trillion this year. Cloud and mobility demands are the biggest drivers for data center systems spending.

Spending on devices is also increasing this year. Spending on IT servers and related equipment has increased to over $600 billion. This year, worldwide spending on IT devices will increase more than 4 percent from 2013. However, organizations will only experience 50 percent utilization of those devices, represents a large capital investment in capacity that is not utilized.

In order to temper the risks of long-term, fixed asset investments and prevent over provisioning, companies must employ smarter forecasting and data center portfolio planning. This can be achieved this by more strategically provisioning facilities. Reducing the cost per kW by just one cent can result in savings of approximately $800,000 per MW over 10 years.

For example, Facebook’s Oregon data center in Prineville is able to operate on 38 percent less power and can complete the same amount of work as the company’s other facilities while costing 24 percent less.

There are a variety of methods organizations can employ to prevent over provisioning in different areas of the data center, including the following:

  • Floor space: In order to reduce over provisioning in this area, operators can turn to modular data center construction or just-in-time strategies.
  • Rack space: Properly planning capacity and utilizing airflow management allows operators to avoid over provisioning.
  • Compute cycle capacity: This is an commonly over provisioned area, but implementing virtualization and application layering enables companies to offset the frequent error.
  • Facility power: As mentioned previously, data center operators can stop over provisioning power use by deploying power management technology and dynamically provisioning computing capacity.
  • Bandwidth: Use of burstable capacity, transit and peering management can help avoid bandwidth over provisioning.

The most reliable way for providers to reduce liability and over provisioning is to spread planning, performance and obsolescence risk across cloud, colocation and data centers.