It is estimated that by 2014, data centers will consume 2% to 3% of the energy produced globally. That is a staggering statistic, especially since the number is on the rise. For the teams that must manage, and pay for, the operation of those data centers, power and cooling costs have become significant portions of the budget.
If you’re like most professionals in the data center world, you’re looking for ways to improve energy efficiency. There are contrasting approaches being taken to achieve this goal, of course. Some IT teams are taking matters into their own hands, deploying meters and logging energy consumption at various times and under various workloads.
However, another option is being touted as an easier, faster path to an understanding of energy in the data center: energy models. The models are delivered in the form of elaborate spreadsheets or interactive software applications, and let data center managers configure their model to match their data center equipment and environment.
It makes sense, right? A model can take into account all of the equipment characteristics, based on detailed vendor specifications, and typical workload applications and let you analyze overall power consumption and associated cooling needs without the time and effort required to measure and consolidate actual power readings.
If these were the only two choices, energy models would be used in every data center today. But the myth of achieving optimal results from a static model is just that — a myth. And fortunately, there is a third alternative that replaces modeled data with real-time, aggregated data from your unique environment. Let’s review how this type of holistic energy management solution dispels the myths surrounding model-based solutions.
Models can provide insights about data center designs and layout and give you an understanding about the factors affecting efficiency at a design level. However, the day-to-day management of energy calls for an accurate view of your dynamic real-time power and temperature levels. And this is where models fall short due to their lack of accuracy.
Models take into account averages and worst-case scenarios. Neither of these will match your unique data center conditions 100% of the time. And models are highly reliant on vendor specifications. Since manufacturers are liable for the accuracy of their published data, they are notoriously conservative in their ratings. Without adjustment, these specifications inevitably lead to over-provisioning, which is exactly what we are all trying to avoid.
Models can provide insights about data center designs and layout and give you an
understanding about the factors affecting efficiency at a design level. However, the day-to-day management of energy calls for an accurate view of your dynamic real-time power and temperature levels. And this is where models fall short due to their lack of accuracy.
Optimistically, we might assume that the developers of model-based energy management solutions take this into account. However, equipment derating is by no means a consistent science. Data center managers, in industry blogs, cite practices of routinely derating by 20% to 50%. And it should also be noted that new equipment draws more power than older equipment. It is naive to assume that models take all of these variations into account since manufacturers do not specify power ratings based on the age of their products. Indeed, real-world measurements confirm that model-based approaches return results that vary significantly when compared to real-time monitoring.
Consider one simple fact: an idle server consumes approximately 60% of its maximum specified power requirements. An energy model cannot help you identify or reduce this major drain on your data center energy budget.
In contrast, a holistic energy management solution gathers and aggregates ongoing power and temperature data throughout your data center. The best solutions gather data that include real-time server inlet temperatures and power consumption data from rack servers and blade servers as well as power-distribution units and uninterrupted power supplies. Airflow is also an important factor that determines energy efficiency in the data center. Monitoring of the airflow parameters along with the temperatures throughout the data center and at the computer-area air handler equipment can also contribute to a better overall picture of the data center energy requirements over time.
By monitoring, collecting, and aggregating all of this power and thermal data, a holistic energy management solution can display a complete picture of the data center activity levels and the related power consumption and temperature levels.
Besides obviously identifying how much time each server is idle, comprehensive monitoring and logging of power and temperature data provides insights into the overall patterns of power consumption in the data center. It enables a knowledge base that allows many improvements to the overall energy efficiency in the data center.
MITIGATING POWER SPIKES
Visibility, for example, arms you, as a data center manager, to understand and mitigate power spikes. Power spikes can lead to serious problems in the data center, and ultimately disrupt services and the business teams that depend on them. They can be the result of the local utility company’s aging equipment, or weather conditions such as heat waves that create havoc when thousands of air conditioning units suddenly drive up power levels on the grid.
To achieve immunity to power spikes, you need to be able to rapidly identify at-risk servers that are operating near their power and temperature limits. Power spike mitigation can focus on those servers. Even better, visibility of server power and temperature conditions puts you in a proactive position. Monitoring over time will allow you to understand the conditions that put equipment and services at risk, so that vulnerability to power spikes can be avoided completely.
Holistic energy management solutions also include automated threshold setting and detection features. Your data center team can uncover hot spots and computer-area air handler failures early before conditions escalate. The best solutions save IT time and simultaneously offer many benefits such as extended life for equipment, avoided outages, and faster risk mitigation in the event of any major power spikes.
Even with the best-possible monitoring and risk mitigation, outages can still occur. The costs of any disaster-related outages can be devastating, adding up to millions of dollars in lost revenue for large enterprises. Besides having disaster recovery plans to safeguard the business during these events, data center managers should be focused on optimizing power utilization during any outage.
Perhaps at no time is the value of real-time monitoring more apparent than during an outage. This is when temperature and power monitoring logs serve as a valuable knowledge base. Armed with accurate power characteristics, IT can intelligently prioritize servers and resources and allocate power accordingly.
Leading-edge energy management solutions also offer power-capping capabilities. Combined with extremely accurate historical power data, IT can appropriately adjust power and server performance. Lower-priority applications can be either disabled or configured to operate at lower performance levels as a method of conserving power. This might be necessary to protect your most vital business processes after switching over to a lower-capacity co-location facility. Power capping can also extend the life of back-up power supplies by up to 25%, based on results from in-field proof-of-concept solution testing.
Unlike model-based energy management solutions, real-time monitoring also enables power management practices that optimize rack densities to stay within power envelopes during normal or restricted levels of operation. This capability not only benefits the data center during outages, but on a day-to-day basis since maximizing rack densities also allows for more energy-efficient cooling system configurations.
REAL-TIME FEEDBACK-BASED CONTROLS
Holistic energy management solutions have been broadly deployed around the world, and the results are extremely compelling. It has been proven that the intelligent aggregation of data center power and thermal data can be used to drive optimal power management policies across servers and storage area networks. In operational data centers, intelligent energy management solutions are producing 20% to 40% reductions in energy waste.
These results are not the sole result of accurate predictions. The state-of-the-art energy management systems provide the ability to monitor, identify patterns and inefficiencies, and dynamically adjust the internal power states of data center servers. It is the addition of these control functions that let IT optimally balance server performance and power.
To support fine-tuning of power consumption in relation to server performance, the leading-edge energy management solutions combine real-time monitoring of actual power consumption with the ability to dynamically adjust the processor operating frequencies. This requires a tightly integrated solution that can interact with the server operating system or hypervisor, based on threshold alerts.
You might be wondering how much you can lower the operating frequencies of a server without lowering its performance. Field tests of state-of-the-art energy management solutions have shown that this type of dynamic adjustment can reduce server power consumption by as much as 20% with negligible impacts on performance.1
Of course, with this level of control, IT can more dramatically adjust power levels when necessary to mitigate power surges or give priority to a mission-critical resource or service. No model-based solution can introduce this level of control without real-time feedback. IT teams might attempt to use models to define energy policies and controls, but a holistic energy management solution uniquely integrates the analysis, control, and mitigation functions with real-time monitoring of data center conditions.
Power is at the heart of optimized resource balancing in the data center, and best-in-class holistic energy management solutions fully exploit control of power. They also introduce more advanced capabilities that further exceed the limited insights that can be gained from a model-centric energy management platform. Adopters are documenting holistic platform use cases that include power-based metering for servers, racks, or rows in the data center, energy charge-back scenarios, and other applications that are motivating conservation and helping IT fairly account for and assign costs to service users.
Basic monitoring and control alone yields excellent ROI for intelligent energy management solutions and the potential for longer-term value makes them a viable investment even during these times of economic recovery. Ultimately, the ability to monitor and manage consumption helps data center operators better define and manage over time the policies and rules for allocating data center power.
The rules and oversight lead to reductions, starting with the identification of under-utilized resources. Numerous field studies have determined that approximately 10% to 15% of all data center servers are idle. Industry analysts have estimated that billions of dollars are spent on server management, power, and cooling for idle servers in the U.S. alone. Other data center managers are using real-time energy monitoring and control platforms to identify opportunities for replacing expensive intelligent power strips with lower-cost alternatives. For a savings of $400 per strip, IT can quickly multiply savings.
An intelligent power management solution can also help IT and facilities evolve a more efficient facility design for cooling and airflow, and can help data center managers make the best possible use of existing floor space and accurately forecast future needs based on expected company growth.
There is no relief in sight when it comes to the cost of energy. Power management has become a business necessity. Before falling for the claims of how a model-based power management solution can help you improve your data center energy efficiency, go back and review all of the advantages of a holistic solution based on the aggregation of real-time data from the data center. Then check out the customer use cases for both types of platforms, and make the best long-term decision for your data center.