Optimizing data center power use is a high priority for data center managers, but they continue to face challenges as power is becoming a larger percentage of ongoing data center costs.
Approximately 10 per cent of data center operating expenditure (opex) is power today, and power is likely to be about 15 per cent of data center opex within five years.
The power usage effectiveness (PUE) metric has become a de facto industry standard for measuring how efficiently a data center uses energy. However, while the PUE metric focuses on the data center energy efficiency of the facilities’ components, it does not provide insights into energy efficiency improvements at the IT equipment level.
In most cases, improving IT equipment power usage can result in the greatest overall improvement in power usage efficiency. There are five key steps data center managers should follow in sequence to maximize efficiency, as well as a ‘Step 0’ – measuring and monitoring power usage – so that improvements can be monitored along the way.
Optimizing IT power
Since IT systems are what ultimately require power delivery, data center managers need to try to lower the amount of power to IT equipment (known as payload power) that’s required. Sixty per cent of the payload power is consumed by servers so taking the following actions to reduce the power they require is crucial:
- Clean up your workloads and eliminate everything that is not necessary
- Consolidate virtual machines
- Virtualize more workloads
- Continue to eliminate those servers that are powered but doing nothing useful
- Replace old servers with newer ones
Optimizing data center space
Data centers built before the advent of server virtualization may be overbuilt for today’s equipment needs, enabling further reduction of the necessary space for IT equipment and less IT power.
When building a new data center, it’s worth considering a modular design that breaks down the data center into individual modules that can be continually refreshed as part of a more flexible and organic data center design.
Optimizing data center cooling
In order to achieve a minimum level of efficiency, data center managers should ensure the adoption of basic data center cooling best practices:
- Install economizers – In cooler regions an air economizer can significantly improve the PUE. In most of North America, for example, 40 to 90 per cent of the cooling can come from outside by using air economizers.
- Contain equipment and heat – Isolation structures can house the data center equipment generating the most heat and funnel the heat out of the data center or to heat other parts of the building.
- Optimize air conditioning systems – There are two primary ways to optimize an air conditioning system – turn it off periodically, using an alternative cooling source, such as an air optimizer or to continually vary the speed, which helps to reduce the total amount of energy the unit consumes.
Eliminating data center power and cooling inefficiencies
Outdated power delivery systems, including uninterruptible power supplies (UPS), power distribution units (PDUs) and transformers, can negatively impact PUE ratios. Evaluating the current situation, future requirements and modern alternatives require time and investments, but typically generate a good return in terms of PUE ratio improvement and savings.
Utilizing DCIM tools
Further refinements to energy efficiency can be achieved through the use of data center infrastructure management (DCIM) software. DCIM software provides the necessary link between the operational needs of the physical IT equipment and the physical facilities (building and environment controls).
Henrique Cecci is a research director at Gartner, Inc. He focuses on enterprise data center strategies and trends, cloud and hosting options, hybrid models, data center futures, power/cooling, green IT, data center design, retrofit, and consolidation and migration strategies.